Molecular switching system using glycosylphosphatidylinositol to select cells highly expressing recombinant proteins

Although many pharmaceutical proteins are produced in mammalian cells, there remains a challenge to select cell lines that express recombinant proteins with high productivity. Since most biopharmaceutical proteins are secreted by cells into the medium, it is difficult to select cell lines that produce large amounts of the target protein. To address this issue, a new protein expression system using the glycosylphosphatidylinositol (GPI)-anchor was developed. PGAP2 is involved in processing GPI-anchored proteins (GPI-APs) during transport. In PGAP2 mutant cells, most GPI-APs are secreted into the medium. Here, we established a HEK293 cell line where endogenous PGAP2 was knocked out and exogenous PGAP2 was inserted with a piggyBac transposon in the genome. Using these cells, human lysosomal acid lipase (LIPA) and α-galactosidase A (GLA) were expressed as GPI-anchored forms (LIPA-GPI and GLA-GPI) and cells expressing high levels of LIPA-GPI or GLA-GPI on the cell surface were enriched. Removal of the PGAP2 gene by piggyBac transposase or FLP recombinase converted LIPA-GPI and GLA-GPI from membrane-bound to the secreted forms. Thus, cells expressing LIPA or GLA in large amounts could be enriched using this approach. The GPI-based molecular switching system is an efficient approach to isolate cells expressing recombinant proteins with high productivity.

. GPI-based protein production system. (A) PGAP2 is required for the addition of the saturated acyl chain to the lyso-GPI moiety, which ensures correct expression of GPI-APs on the cell surface. In PGAP2 mutant cells, lyso-GPI-APs are transported to the cell surface and then most of the GPI-APs are secreted into the medium. Since lyso-forms of GPI-APs are very sensitive to PLD, the GPI structure of secreted proteins become PLD-cleaved forms. (B) The GPI-anchoring system ensures the expression of target proteins as GPI-APs on the cell surface. Cells expressing target proteins at a high level can be enriched by selection of the GPIanchored target proteins. Removal of the PGAP2 gene results in higher amounts of the target proteins being secreted into the medium. present in the cell population that are producing high levels of recombinant protein, it is often challenging to select such a high-producing cell line from the bulk cell population. To overcome this issue, we tried to develop a new system to select cells with high protein productivity from a mixture of cells. In this system, we used GPI anchoring of proteins (Fig. 1A). When the secretory target protein is expressed using this system, the proteins are expressed as GPI-anchored forms (Fig. 1B). Therefore, cells expressing high amounts of a target protein can be enriched by sorting the cells for the GPI-anchored form. Finally, the GPI-anchored membrane-bound form can be switched to the secretory form by disruption of the GPI biosynthetic pathway.
To achieve this disruption, the PGAP2 gene in HEK293 cells was knocked out using CRISPR/Cas9. PGAP2 is required for the addition of a saturated fatty acid to lyso-GPI-anchors during the GPI fatty acid remodelling reaction in the Golgi (Fig. 1A) 17 . Deficiency of this protein results in the transport of lyso-GPI-APs, which are sensitive to PLD. Consequently, significant parts of GPI-APs are secreted into the medium 17 . As expected, when the PGAP2 gene was knocked out in HEK293 cells, surface expression of CD59, a ubiquitously expressed GPI-AP, was decreased significantly (Fig. 2A). The KO target region was amplified from the genome and sequenced, showing that there was the 66 bp deletion between target 1 and 2 sequences in the coding sequence (Fig. 2B,C). The deletion causes removal of 22 amino acids from the luminal region of the PGAP2, suggesting that this mutation inactivates protein function. Although the mutation is associated with hyperphosphatasia with mental retardation syndrome 24,25 , PGAP2-KO cells did not show any growth and shape defects at the cellular level.
Rescue of the PGAP2 gene using PB-mediated insertion to the PGAP2-KO cells. The piggy-Bac (PB) transposon system allows reversible transgenesis and precise re-excision from the genome 26,27 . The PGAP2 gene was rescued by a plasmid carrying the PB transposon. The plasmid pPB-FRT-PuroΔTK-PGAP2 construct includes the PGAP2 gene and the puroΔTK selection cassette flanked by PB and FRT sequences (Fig. 3A). PGAP2-KO cells were co-transfected with pCMV-hyPBase expressing the PB transposase (PBase) and pPB-FRT-PuroΔTK-PGAP2 to integrate the PGAP2 gene into the genome (Fig. 3A). After transfected cells were treated with puromycin, surviving cells were diluted and clonal cell lines were obtained. When the PB transposon is used for gene integration into the genome, several copies of the construct may be found in a cell 28 . However, a cell line that contains only one PB copy number in the genome was desired for further purposes. Among the clonal cells, we chose one cell line, named HEK293pB-PGAP2 clone P15 (P15), which rescued the expression of CD59 on the cell surface and had only one PB insertion site in the genome (Fig. 3B,C). The PB element containing the PGAP2 gene was inserted at the TTAA sequence of the intron of the SBF2 (SET Binding Factor 2) gene located on chromosome 11 (Fig. 3D).
Since the integrated PGAP2 fragment is flanked by FRT sites and the PB transposon at both ends, the PGAP2 gene can be re-excised by FLP recombinase or PBase (Fig. 4A). In addition, puroΔTK was used as the positive and negative selection marker for insertion. The cells excised with the PB element containing PGAP2 and puroΔTK genes are resistant to treatment with the nucleoside analogue, 1-(2-deoxy-2-fluoro-1-Darabinofuranosyl)-5-iodouracil (FIAU). After the plasmids expressing FLP recombinase or PBase were transfected into P15 cells, followed by FIAU treatment, the surface expression of CD59 on the surviving cells was analysed by flow cytometry. By expression of PBase or FLP recombinase, the P15 phenotype was reverted to that of PGAP2-KO cells, suggesting that the PGAP2 gene was excised correctly (Fig. 4B). These data indicate that the P15 cell line can easily convert from GPI-positive cells to GPI-less cells by expression of PBase or FLP recombinase. By deletion of PGAP2, GPI-APs can be switched from membrane-bound forms to secreted forms.

Expression of recombinant proteins using GPI anchors.
Recombinant proteins can be expressed as GPI-APs by simply conferring a GPI-attachment signal sequence to the carboxyl-terminus of the target protein.
In the ER, the carboxyl-terminal GPI-attachment signal is exchanged with preformed GPI and this reaction is mediated by GPI transamidase 10 . As a model protein to express in this system, we chose lysosomal acid lipase (LIPA), which is an enzyme required for breakdown of lipids such as cholesterol esters and triacylglycerols in lysosomes 29 . Mutations in the LIPA gene cause lysosomal storage diseases such as Wolman disease and cholesteryl ester storage disease 30 . For treatment of LIPA deficiency, Kanuma (Sebelipase alfa), which is a recombinant LIPA produced by transgenic chicken egg, was approved for use in 2015 in the USA and EU 31 . We constructed vectors carrying His6-Flag-tagged LIPA (sHF-LIPA) or sHF-LIPA fused with a GPI-attachment signal (sHF-LIPA-GPI), which were stably expressed in cells (Fig. 5A). The sHF-LIPA-GPI was expressed on the cell surface and cleaved by treatment with PI-PLC (Fig. 5B), indicating that LIPA was expressed as a GPI-anchored form.

Selection of cells highly expressing recombinant proteins.
The developed over-expression system enabled the sorting of cells expressing high amounts of LIPA. The plasmid for the expression of sHF-LIPA-GPI was transfected into P15 cells and the cells stably expressing sHF-LIPA-GPI were selected with antibiotics. The cells with high levels of expression of sHF-LIPA-GPI on the cell surface were then enriched using a cell sorter (Fig. 6A). After sorting twice, the expression of sHF-LIPA-GPI on the cell surface was increased by 28-fold (with the geometric mean fluorescent intensity) when compared with that of unsorted cells. The PGAP2 gene was then removed from the sorted cells by expression of PBase or FLP recombinase. The surface expression of CD59 and sHF-LIPA-GPI decreased (Fig. 6B). Removal of the PGAP2 gene from cells led to the secretion of sHF-LIPA into the medium (Fig. 6C), whereas P15 cells expressed sHF-LIPA-GPI on the cell surface but did not secrete this protein into the medium. Compared with cells that were sorted once, higher amounts of sHF-LIPA were secreted into the medium for cells that were sorted twice, suggesting that the system requires two sorting steps to ensure maximum selection and thus protein yields. LIPA activity was detected from the culture media of wild-type HEK293 cells and P15 cells expressing soluble sHF-LIPA (Fig. 7A). In contrast, the activity was not detected in the medium of P15 cells expressing sHF-LIPA-GPI without treatment. Once PGAP2 was removed by either PBase or FLP recombinase, however, a significant level of LIPA activity was detected (Fig. 7A). The activities from the first sorted cells and the second sorted cells were 1.8 and 2.0 times higher than that of P15 cells expressing soluble sHF-LIPA, respectively.
After the second sorting, we isolated clonal cell lines from P15 cells expressing sHF-LIPA-GPI by limiting dilution. We selected 10 clonal cell lines from P15 cells expressing sHF-LIPA and P15 cells expressing sHF-LIPA-GPI that were sorted twice followed by the removal of PGAP2. The sHF-LIPA was expressed in the medium at higher amounts in cells using the GPI-anchoring system when compared with that of the control P15 cells stably expressing soluble sHF-LIPA (Fig. 7B). We picked up one clonal cell line No. 5 that showed highest LIPA activity and detect the LIPA proteins. The protein levels secreted from clone No. 5 were increased 1.7 times compared to those from wild type cells expressing soluble LIPA (Fig. 7C). Therefore, the system can better and more easily select cells that are expressing recombinant proteins when compared with previous methods of secreting proteins into the medium.
High expression of α-galactosidase A using GPI-based protein expression system. We finally analysed whether the system can be applied to express other recombinant proteins. As the protein to express, we chose the α-galactosidase A (GLA), which is another lysosomal enzyme. Recombinant GLA is used for the enzyme replacement therapy of the Fabry diseases 32 . Plasmids expressing His6-Flag-tagged GLA (sHF-GLA) and GPI-anchored sHF-GLA (sHF-GLA-GPI) were constructed and transfected into P15 cells (Fig. 8A). After selection by antibiotics, P15 cells expressing sHF-GLA-GPI were stained with anti-Flag antibodies and cells highly expressing GPI-anchored sHF-GLA on the cell surface were enriched twice by cell sorter. The surface level of sHF-GLA was increased 3.2 times (Fig. 8B). When PGAP2 gene was removed from cells by the expression of FLPe recombinase, surface levels of sHF-GLA was decreased. Instead, the proteins started being released into the medium (Fig. 8C). The secreted GLA from cells sorted once and twice were higher than those from unsorted cells and cells expressing soluble sHF-GLA. The activity of GLA was increased in medium prepared from cells sorted twice (Fig. 8D). These results indicate that the system is useful to select cells highly expressing recombinant proteins of interest.

Discussion
In this study, we developed a new system to select cells expressing high amounts of recombinant proteins. The advantage offered by GPI-anchoring of proteins was used to tether proteins to the cell surface 11,19 . By enrichment  line, named P15, in which GPI-APs can be converted from membrane-bound forms to secretory forms. Initially, the PGAP2 gene was knocked out in HEK293 cells to secrete GPI-APs into the medium. Second, the PGAP2 gene was rescued in KO cells by a PB transposon-based plasmid. By introduction of PB transposase or FLP recombinase, the PGAP2 gene can be easily removed from P15 cells, yielding PGAP2-KO cells once more. Using the P15 cells, recombinant proteins, LIPA and GLA, were expressed as the GPI-anchored form. After sorting cells that expressed high levels of GPI-anchored forms of proteins on the cell surface, LIPA-GPI and GLA-GPI were switched from the membrane-bound form to the secretory form by removal of the PGAP2 gene from the cells. The system does not require treatment with any GPI-cleaving enzymes for release. Lyso-forms of GPI-APs are further cleaved by PLD-like activity 17,18 . The activity arises from both endogenous expression in cells and the serum. With this method, we obtained cell lines that expressed recombinant LIPA and GLA at higher levels when compared with conventional methods.
When a GPI-attachment signal is added to the C-terminus of secretory proteins, the proteins are expressed as GPI-APs. Therefore, it has been possible to express a wide range of recombinant proteins on the cell surface through GPI-anchors 19,33 . In particular, it would be useful to tether soluble proteins and type-I membrane proteins. GPI-anchored LIPA and GLA were functional, thereby showing that the GPI moiety did not hinder proper protein folding and function. Here, we used His6-Flag-tagged LIPA and GLA as model recombinant proteins to express, because such tags are useful for protein purification and detection. The system can also be used with non-tagged native proteins if an antibody against the target protein is available. In our system, target proteins expressed on the cell surface are stained with antibodies and the high expressing cells are enriched using flow cytometry-based cell sorters or magnetic separators. Therefore, it is a requirement that antibodies can be used with flow cytometry.
In this system, the GPI-glycan part is attached to the C-terminus of proteins, which has both advantages and disadvantages with respect to affecting the function and properties of the target protein. A clear disadvantage is that C-terminal labelling may lead to loss of protein function, and thus our system cannot be used for proteins that suffer loss of activity upon C-terminal modification 34 . In addition, because proteins are modified with the GPI-glycan, they are not identical to their native forms. Since soluble GPI-APs, which are cleaved by GPI-cleaving enzymes, exist in our body, cleaved GPI itself would not become the antigen. When the proteins are used for clinical purpose, however, attention should be paid to immunogenicity, because attachment of the GPI-glycan yields non-native target proteins. These negative points need to be addressed and improved in future studies.
Advantages of the expression system include protection from carboxypeptidases degradation because of C-terminal modification with a glycan moiety 35 . Second, there are several proteins that recognize GPI-glycans such as α-toxin and aerolysin 36,37 , which can be used for enrichment and purification of target proteins 38 . Third, GPI-anchors have unique structures and are chemically and enzymatically processed 39,40 . GPI moieties are also chemically modified in vivo [41][42][43] . Therefore, the GPI-anchored recombinant proteins could be labelled or crosslinked with reagents such as fluorescent probes and nanoparticles. Further research would help to expand the applications of the system.
Plasmids expressing target proteins with monitoring proteins such as GFP are used for the selection of cells expressing recombinant proteins in high levels 44,45 . In these methods, it is assumed that cells highly expressing monitoring proteins are also able to secrete large amounts of recombinant proteins. Since our method monitors the recombinant protein on the cell surface, it is a more direct approach to monitor protein expression levels. The dihydrofolate reductase (DHFR)-methotrexate (MTX) system is frequently used for increasing the expression of recombinant pharmaceutical proteins 46 . In this system, plasmids carrying a target gene of interest with DHFR are transfected into dhfr-deficient cells. By increasing the DHFR inhibitor MTX to select cells, the DHFR region together with target gene are amplified in the genome. Our system can be combined with the DHFR system to improve protein productivity. The advantage of our system is that cells expressing high levels of the target recombinant protein are enriched by cell sorting. Our findings show that the GPI-based protein expression system is more efficient when compared with other methods, as it is faster and highly productive. The system should be useful for large-scale production of recombinant proteins.
Plasmid construction. The plasmid pME-Puro-sHF-GPI containing a signal sequence, a His6 tag, a Flag tag and a GPI attachment signal of human CD59 were used to construct a plasmid for expression of GPI-anchored LIPA and GLA. LIPA and GLA fragments were obtained from human cDNA amplified using the primer sets 5′-CGATGACAAGCTCGAGGGGAAACTGACAGCTTTG-3 ′ / 5 ′ -T C

C C A C C A T T C T C G A G C T G A TA T T T C C T C A T TA G A T TA A TA A T T -3 ′ a n d 5 ′ -C G AT G A C A A G C T C G A G C T G G A C A AT G G AT T G G C A A G G -3 ′ / 5 ′ -T C C C A C C AT T
CTCGAGAAGTAAGTCTTTTAATGACATCTGCATTG-3′, respectively. The fragments were then ligated into the XhoI site of pME-Puro-sHF-GPI by In-fusion cloning (Takara, Shiga, Japan). The resulting constructs, pME-Puro-sHF-LIPA-GPI and pME-Puro-sHF-GLA-GPI, were confirmed by sequencing. The sHF-LIPA-GPI and sHF-GLA-GPI fragments were digested and ligated into the pME-Hyg plasmid to yield the pME-Hyg-sHF-LIPA-GPI and pME-Hyg-sHF-GLA-GPI plasmids, respectively.
For the plasmid expressing the soluble form of LIPA or GLA, a DNA fragment corresponding to the mature LIPA or GLA sequence after the ER insertional sequence was amplified and ligated into XhoI and NotI sites of pME-Hyg-sHF-LIPA-GPI to generate pME-Hyg-sHF-LIPA and pME-Hyg-sHF-GLA. For gene rescue mediated by the PB system, pPB-FRT-PGKp-PuroΔTK was constructed and contained a PGK promoter, a multiple cloning site, a bovine growth hormone (GH) poly-adenylation signal, SV40 promoter and puroΔTK flanked by both PB terminal repeat sequences and flippase recognition target (FRT) at both ends. The Flag-tagged rat PGAP2 gene was amplified and cloned into EcoRI and NotI of pPB-FRT-PGKp-PuroΔTK to give pPB-FRT-PuroΔTK-PGAP2. The pCMV-hyPBase was kindly provided from Kosuke Yusa (Sanger Institute) and pCAG-FLPe-IRES-puro was purchased from Addgene.
Flow cytometry and cell sorting. Cells were harvested by a trypsin/EDTA solution for CD59 staining or by PBS containing 2 mM EDTA and 0.5% BSA for Flag staining. Cells (5 × 10 5 )/sample were stained with anti-CD59 or anti-Flag (10 μg/ml) as the first antibody, and PE-conjugated goat anti-mouse IgG was used as the second antibody. After staining, cells were analysed using Accuri C6 (BD, San Jose, CA, USA). Where necessary, cells were treated with phosphatidylinositol-specific phospholipase C (PI-PLC: Thermo Scientific, Carlsbad, CA, USA) for 1.5 h at 37 °C prior to antibody staining. For cell sorting, the cell sorting solution (PBS, 2 mM EDTA and 0.5% BSA) was used for resuspension of antibodies and washing instead of the FACS solution. Stained cells were sorted by an S3e Cell Sorter. PB copy number determination. To rescue PGAP2-KO cells, pPB-FRT-PuroΔTK-PGAP2 was co-transfected with pCMV-hyPBase. Two days after transfection, the cell medium was replaced by medium containing puromycin to select positive cells. After the limiting dilution, inserted copy number and insertion sites of the PB-FRT-PuroΔTK-PGAP2 fragment were analysed for the obtained clones. The genomic DNAs from each clone were isolated using the Wizard genomic DNA purification kit (Promega, Madison, WI, USA). Genomic DNA was digested by HaeIII, followed by ligation of adaptors. Splinkerette-PCR was performed to determine PB insertion sites 50,51 . Flanking sequences were amplified using primer sets Spl-P1, 5′-CGAATCGTAACCGTTCGTACGAGAA-3′ and 3PB-1st, 5′-TATACAGACCGATAAAACACATGCGT -3′ for the 1st PCR, and Spl-P2 5′-TCGTACGAGAATCGCTGTCCTCTCC-3′ and 3PB-2nd, 5′-CGCATGATTATCTTTAACGTACGTCACAA-3′ for the second nested PCR. To determine PB copy number, amplified DNA was applied on 2% agarose gel and checked by electrophoresis. The insertion sites were determined by sequencing. The cell line, named P15, with a single PB copy in the genome was chosen for further experiments.
Recombinant mammalian protein expression. To make stable cell lines expressing soluble forms sHF-LIPA or sHF-GLA and membrane-bound forms sHF-LIPA-GPI or sHF-GLA-GPI, linearized DNA plasmids were transfected into cells. Two days after transfection, cells were selected by culturing in medium containing hygromycin (400 μg/ml). To remove PB transposons from cells expressing GPI-fused proteins, sorted cells in a 6-well plate were transfected with 4 μg DNA of pCMV-hyPBase or pCAG-FLPe-IRES-puro for PBase or Flippase, respectively. After 2 d, medium containing 1 μM FIAU was applied, and the medium was changed every day for 7 d.
Western blotting of sHF-LIPA and sHF-GLA prepared from medium and cell lysates. Cells (5 × 10 5 ) were plated and cultured in 2 ml of medium. After 48 h, 1.4 ml medium was collected. The medium was centrifuged at 10,000 × g for 5 min and 1000 μl of the supernatant was collected in a new tube. Then, 20 μl of prewashed anti-Flag beads (MBL, Aichi, Japan) for DDDDK-tagged proteins were added. The tube was rotated at 4 °C for 2 h, and then centrifuged at 10,000 × g for 1 min. The supernatant was removed and the beads with tagged proteins were washed 3 times with cold PBS. The proteins were eluted with the Flag peptide. For the cell lysates, after removing the medium, 5 × 10 5 cells were harvested with the cell sorting solution and washed using cold PBS at 4 °C. One hundred microliters of lysis buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% Triton-X100, 1 mM EDTA, protein inhibitor, 1 mM PMSF) was added to the cell pellets and the mixture was incubated on ice for 30 min. After incubation, the tube was centrifuged at 10,000 × g for 15 min at 4 °C. The supernatant was collected and mixed with sample buffer and boiled at 95 °C for 5 min. The solution was kept at −20 °C until western blotting. The recombinant LIPA and GLA were detected using the anti-Flag antibody and horseradish peroxidase (HRP)-conjugated anti-mouse IgG as the primary antibody and secondary antibody, respectively. Proteins were visualized using ImageQuant TM LAS 4000 (GE Healthcare, Little Chalfont, UK). LIPA activity. The medium was collected after culturing 2.5 × 10 5 cells in a 12-well plate for 48 h in 1 ml medium without any antibiotics. After centrifugation at 10,000 × g, 0.8 ml of medium was collected and then stored at −20 °C until use. LIPA activity was measured by an optimized assay for lipase activity using cell medium 52,53 . Briefly, 100 ml assay buffer (100 mM citrate buffer (pH 5.0), 0.5% Triton X-100, 1 mM CaCl 2 ) and 20 μM 4-nitrophenyl-palmitate (4-NPP) dissolved in isopropanol/acetonitrile (4/1 v/v) were prepared before the assay. The assay mixture was obtained by adding 4-NPP to the assay buffer to give a final concentration of 4-NPP of 1 μM. The mixture was then placed in a 60 °C water bath for 5 min, mixing continuously until the solution was transparent. An appropriate amount of medium was centrifuged at 10,000 × g at 4 °C for 5 min, and the supernatant was collected in a new tube. Ten microliters of cell medium was added to 60 μl assay mixture in transparent 96-well plates, which was immediately covered by a dark paraffin paper before placing it in the incubator at 30 °C for 45 min. Subsequently, 120 μl of 1 M Tris-HCl (pH 8.0) was added to the assay mixture to terminate the reaction. The absorbance was measured at 405 nm, using an Enspire 2300 Multilabel Reader (Perkin Elmer, Waltham, MA. USA). One unit of lipase activity was defined as the amount of 4-nitrophenol nmol per min released from 4-NPP at 30 °C. GLA activity. The medium for GLA activity was prepared the same as described above for LIPA activity.