Double knockdown of α1,6-fucosyltransferase (FUT8) and GDP-mannose 4,6-dehydratase (GMD) in antibody-producing cells: a new strategy for generating fully non-fucosylated therapeutic antibodies with enhanced ADCC

Background Antibody-dependent cellular cytotoxicity (ADCC) is greatly enhanced by the absence of the core fucose of oligosaccharides attached to the Fc, and is closely related to the clinical efficacy of anticancer activity in humans in vivo. Unfortunately, all licensed therapeutic antibodies and almost all currently-developed therapeutic antibodies are heavily fucosylated and fail to optimize ADCC, which leads to a large dose requirement at a very high cost for the administration of antibody therapy to cancer patients. In this study, we explored the possibility of converting already-established antibody-producing cells to cells that produce antibodies fully lacking core fucosylation in order to facilitate the rapid development of next-generation therapeutic antibodies. Results Firstly, loss-of-function analyses using small interfering RNAs (siRNAs) against the three key genes involved in oligosaccharide fucose modification, i.e. α1,6-fucosyltransferase (FUT8), GDP-mannose 4,6-dehydratase (GMD), and GDP-fucose transporter (GFT), revealed that single-gene knockdown of each target was insufficient to completely defucosylate the products in antibody-producing cells, even though the most effective siRNA (>90% depression of the target mRNA) was employed. Interestingly, beyond our expectations, synergistic effects of FUT8 and GMD siRNAs on the reduction in fucosylation were observed, but not when these were used in combination with GFT siRNA. Secondly, we successfully developed an effective short hairpin siRNA tandem expression vector that facilitated the double knockdown of FUT8 and GMD, and we converted antibody-producing Chinese hamster ovary (CHO) cells to fully non-fucosylated antibody producers within two months, and with high converting frequency. Finally, the stable manufacture of fully non-fucosylated antibodies with enhanced ADCC was confirmed using the converted cells in serum-free fed-batch culture. Conclusion Our results suggest that FUT8 and GMD collaborate synergistically in the process of intracellular oligosaccharide fucosylation. We also demonstrated that double knockdown of FUT8 and GMD in antibody-producing cells could serve as a new strategy for producing next-generation therapeutic antibodies fully lacking core fucosylation and with enhanced ADCC. This approach offers tremendous cost- and time-sparing advantages for the development of next-generation therapeutic antibodies.


Background
Antibodies of the human IgG1 isotype containing two biantennary complex-type N-linked oligosaccharides in the constant region (Fc) [1] are commonly used therapeutically. As regards cancer treatment in particular, the antibody effector function of antibody-dependent cellular cytotoxicity (ADCC) is known to be important and is closely related to clinical efficacy in humans in vivo [2][3][4]. Through the Fc, therapeutic antibodies can mediate effector functions, and ADCC is greatly influenced by Fc oligosaccharide structure [5,6]. Removal of the core fucose from Fc oligosaccharides is widely recognized as being important for the effector function of ADCC [7,8]. Antibodies in which the Fc oligosaccharide structure lacks the core fucose exhibit more potent efficacy than do fucosylated antibodies, both in vitro and in vivo [9][10][11][12][13]. Therapeutic antibodies fully lacking core fucosylation are able to escape the inhibitory effects of both human serum IgG and other contaminating fucosylated antibody ingredients to achieve optimal ADCC [6,[14][15][16][17]. Unfortunately, almost all licensed therapeutic antibodies developed to date are heavily fucosylated, i.e., the majority of antibody molecules possess Fc oligosaccharides with the core fucose [18,19], which results in a failure to optimize ADCC. The presence of this core fucose is largely due to the fact that the antibodies are produced by rodent mammalian cell lines with intrinsic fucosyltransferase activity (e.g., Chinese hamster ovary (CHO), mouse myeloma NS0 and SP2/0, and mouse hybridoma cell lines).
In mammalian cells, core fucosylation of the Fc oligosaccharides is mediated by the only gene, α1,6-fucosyltransferase (FUT8), that catalyzes the transfer of fucose from GDP-fucose to the innermost N-acetylglucosamine (Glc-NAc) of Fc oligosaccharides via an α1,6-linkage [20]. The intracellular GDP-fucose, an essential substrate of oligosaccharide fucosylation, is synthesized in the cytoplasm via both a de novo pathway and the salvage pathway shown in Fig. 1. The de novo pathway transforms GDPmannose, which originates from D-glucose taken into the Oligosaccharide fucosylation and GDP-fucose synthesis in mammalian cells *O\FRSURWHLQ cytoplasm from the extracellular environment, to GDPfucose, via three enzymatic reactions carried out by two proteins: GDP-mannose 4,6-dehydratase (GMD) and GDP-keo-6-deoxymannose 3,5-epimerase, 4-reductase (FX) [21,22]. The salvage pathway synthesizes GDPfucose from free L-fucose derived from extracellular or lysosomal sources. Most of the intracellular GDP-fucose is generated via the de novo pathway, and the metabolite-free L-fucose is also reutilized through the salvage pathway [22]. The GDP-fucose, which accumulates in the cytoplasm, is transported into the lumen of the Golgi apparatus by a GDP-fucose transporter (GFT) anchored at the Golgi membrane [23], and then serves as a substrate in the synthesis of fucosylated glycoconjugates by fucosyltransferases [22,24,25].
To date, only a few studies have addressed the regulation of Fc oligosaccharide fucosylation in mammalian cells using the following approaches: 1) the application of a mutant CHO cell line, Lec13, partially deficient in GMD [7] or that of a rat hybridoma cell line, YB2/0 [8], as host cells; 2) the introduction of a small interfering RNA (siRNA) against FUT8 [26]; and 3) the co-expression of β-1,4-N-acetylglucosaminyltransferase III (GnT-III) and Golgi α-mannosidase II (ManII) [27]. Of these, the gene knockout of FUT8 and GMD is the only strategy for manufacturing fully non-fucosylated recombinant therapeutics in mammalian cells [28][29][30]. However, gene targeting in mammalian somatic cells is difficult to achieve and is a laborious and time-consuming process, because in somatic cells, non-homologous recombination events occur several orders of magnitude more frequently than homologous recombination [28,31]. Thus, gene targeting in mammalian somatic cells remains very difficult to apply to each antibody-producing clone as a simple means of controlling fucosylation.
In this study, we explored the possibility of the high-frequency conversion of already-established antibody-producing cells (1 st -generation) to cells that produce antibodies fully lacking core fucosylation (2 nd -generation) within a few months. Moreover, it was considered industrially useful to generate 2 nd -generation cells that would yield the equivalent antibody productivity of the original cells, without inducing any changes in cell character beyond the lack of fucosylation. To this end, we applied an RNA interference (RNAi) technique in combination with a cellular phenotypic selection strategy using Lens culinaris agglutinin (LCA) lectin, which recognizes the α1,6 fucosylated trimannose-core structure of Nlinked oligosaccharides and commits cells expressing this structure to a cell-death pathway. In our previous study, single-gene knockdown of FUT8 resulted in a substantial, but not full, reduction of antibody fucosylation in the products [26]. Here, we identified synergistic effects of the double knockdown of FUT8 and GMD on oligosaccharide fucosylation in mammalian cells. This finding enabled us to design a new conversion strategy for the manufacture of next-generation therapeutic antibodies fully lacking core fucosylation and with enhanced ADCC.

Single gene knockdown of FUT8, GMD, and GFT in antibody-producing cells
Three constitutive short hairpin siRNA expression vectors against Chinese hamster FUT8, GMD, and GFT were generated and introduced into an IgG1 antibody-producing CHO/DG44 clone, 32-05-12 [26], to evaluate the effects of target gene knockdown on the levels of fucosylation of the products. Puromycin-resistant clones transformed by the vectors appeared after transfection with a transformation efficiency of approximately 1,300 per 1.6 × 10 6 electroporated cells, irrespective of the vector differences. However, subsequent selection of LCA-resistant clones showed clear differences in the appearance of surviving colonies. The ratios of LCA-resistant clones to puromycinresistant clones transformed by the three siRNA expression vectors against FUT8, GMD, and GFT were 11.5%, 14.8%, and 0.8%, respectively ( Table 1). The resultant LCA-resistant clones were expanded in medium without LCA, and the mRNA expression of FUT8, GMD, and GFT, as well as the oligosaccharide structure of the antibodies produced in these clones, were quantitatively analyzed. The target gene mRNA was reduced to a level exceeding 90% in clones appearing in all three siRNA-introduced transformants; however, in none of the clones was Fc oligosaccharide fucosylation completely abolished. The most effective rates of reduction in fucosylation among

Synergistic effects of double knockdown of FUT8 and GMD on fucosylation
To explore the effects of double knockdown of the target genes FUT8 and GMD, two GMD siRNA-introduced clones exhibiting a maximum reduction in product fucosylation, designated as GMDb2 and GMDb5, were selected and re-electroporated with the FUT8 siRNA expression vector. LCA-resistant clones were selected in the presence of 100 μM L-fucose after drug-resistant selection had been carried out, and two clones expressing highly non-fucosylated antibodies were selected independently from each GMD siRNA-introduced clone in the medium. In the four established clones, the FUT8 mRNA expression levels decreased to roughly one-tenth of the parental expression levels, while the original 90% decrease in GMD mRNA was retained. Monosaccharide composition analysis showed that the ratio of non-fucosylated oligosaccharides of the antibodies produced by the four clones had increased significantly in each case (i.e., up to 91%, 95%, 97%, and 98%), compared to those of parental GMD siRNA-introduced clones ( Table 2). The increase in non-fucosylated product levels was cancelled when the clones were cultured in the presence of L-fucose; this cancellation was ascribed to the absence of a GMD knockdown effect due to activation of the salvage pathway of GDP-fucose synthesis. No significant effect on fucosylation levels in antibody-producing cells was observed with a combination of either FUT8 and GFT or GMD and GFT double knockdown.

Conversion of fucosylated antibody-producing cells to non-fucosylated antibody producers
To facilitate the double knockdown of FUT8 and GMD, the siRNA tandem expression vector for targeting these genes was generated and introduced into IgG1 antibodyproducing CHO/DG44 32-05-12 cells. Hygromycinresistant clones transformed by the vectors appeared with a common transformation efficiency of approximately 1,500 per 1.6 × 10 6 electroporated cells. However, subsequent selection of LCA-resistant clones revealed clear differences in the appearance of surviving colonies, as compared to that of single siRNA-introduced clones ( Table 1). The ratio of LCA-resistant clones to hygromycin-resistant clones transformed by the siRNA tandem expression vector was much higher (more than triple) than that of clones transformed by the vector bearing each single siRNA expression cassette; approximately half of the hygromycin-resistant clones showed resistance to LCA, even in the culture condition containing L-fucose. Among the resultant LCA-resistant clones, twenty were randomly selected and expanded in medium lacking both LCA and L-fucose. The mRNA expression of both target genes, FUT8 and GMD, in each clone decreased to approximately less than 20% of that of the parental clone ( Fig.  2A). Out of twenty clones, eight exhibited relatively low levels of both target genes, and these were selected for further analysis of the oligosaccharide structure of the antibody products. The Fc oligosaccharide structures of the antibodies produced by these eight clones showed almost complete non-fucosylation; there were no detectable Lfucose residues among the products from six clones ( Table  3). These six clones exhibited very low reactivity to LCA, even after adaptation to serum-free medium, as was also observed in the case of the FUT8-knockout CHO cell line Ms705. The results obtained from two representative clones, designated as FG1 and FG16, are shown in Fig.  2B-2E.

Serum-free fed-batch culture of siRNA-introduced cells producing non-fucosylated antibodies
Serum-free fed-batch culture of the clones transformed by the FUT8 and GMD siRNA tandem expression vector was carried out using 1L-scale spinner bioreactors with pH and DO controls. Two FUT8 and GMD siRNA-introduced clones, FG1 and FG16, and their parental IgG1-producing clone, CHO/DG44 32-05-12, were adapted to serum-free medium, and the performance of the fed-batch cultures was compared in a head-to-head analysis (Fig. 3A, 3B). The fed-batch cultures were maintained until the cell viability decreased to less than 50%, which occurred at day 16 post-inoculation. Culture aliquots were taken at days 3, 6, 9, 12, 14, and 16 to analyze the cells and antibody products. Two siRNA-introduced cell lines grew logarithmically, with a slight difference in the specific production rate, maximum viable cell density, and the day upon which the viable cell density reached the maximum level; Analyses of clones transformed by the FUT8 and GMD siRNA tandem expression vector two siRNA-introduced hygromycin-resistant descendants showed slightly slower growth and death rates than those of the parent cells. However, the productivity of two clones transformed by FUT8 and GMD siRNAs reached a production level of approximately 1200 mg/L, which was basically equivalent to that of the parental cells at the end of the fed-batch culture period. During the culture period, the levels of cellular FUT8 and GMD mRNAs in the two siRNA-introduced clones remained at the original low level, i.e., less than 20% of the parental level. Monosaccharide composition analysis revealed that the two siRNAintroduced clones stably produced non-fucosylated antibodies throughout the culture period ( Table 4). The oligosaccharide profile analysis of products purified from the final culture medium confirmed that the N-linked Fc oligosaccharides of the products were of the biantennary complex type, and the products from the two siRNA-introduced clones fully lacked core fucosylation (Fig. 3D, 3E). It was also confirmed that there was no significant change in the oligosaccharide profile during the culture period, and the non-fucosylated products showed two orders of magnitude higher ADCC than the antibodies from the parental CHO/DG44 cells, without any changes in antigen binding (Fig. 4).

Discussion
The conversion of fucosylated therapeutic antibodies to antibodies lacking the core fucose in the Fc is recognized as an attractive approach for generating next-generation therapeutics. In this study, we explored the possibility of the high-frequency conversion of already-established antibody-producing cells (1 st -generation) to cells that produce antibodies fully lacking core fucosylation (2 nd -generation) within a few months, while almost all original features of the 1 st -generation antibody-producing cells, including cell growth and recombinant protein productiv-ity, were retained. The present approach is expected to be industrially useful for the establishment of a master cell bank of antibody-producing cells for the rapid development of 2 nd -generation therapeutic antibodies.
Here, we focused on three key genes involved in the oligosaccharide fucosylation pathway in mammalian cells (FUT8, GMD, and GFT), and we identified an effective siRNA of each gene among more than ten siRNA candidates using a transient expression system with a target gene-GFP fusion reporter construct, as previously reported (data not shown) [32]. Every identified siRNA showed high efficacy in terms of depressing the target mRNA to levels of over 90%, when it was expressed in the form of short hairpin siRNA under the control of U6 or the tRNAval promoter in antibody-producing cells in which the siRNA expression unit was integrated in the genome. The present results thus demonstrated that our established siRNA expression system was able to sufficiently decrease level of target mRNA to the current maximum level achievable with RNAi technology. The levels of reduction in Fc fucosylation among the products were found to vary, depending on both target gene and clone, even when specific target genes were reduced to comparable levels in the siRNA-introduced cells. Cellular phenotypic selection of resistance to LCA was conducted to enrich transformants with cellular features equivalent to those of FUT8-knockout cells. However, among clones transformed by either siRNA alone, none produced fully lacking the core fucose of the Fc. These results suggest that there is a fucosylationreducing limit associated with single-gene knockdown of FUT8, GMD, and GFT using RNAi technology.
On the other hand, we clearly observed a noteworthy synergistic effect of double knockdown of FUT8 and GMD on the fucosylation of products in antibody-producing cells ( Table 2), although neither the combination of FUT8 and GFT, nor GMD and GFT double knockdown caused a significant change. This synergistic effect was not accounted for by the levels of mRNA regulation of target genes, because no additive effect on the level of depression of each target mRNA was observed with the FUT8 and GMD double knockdown. Some previous studies have shown that there is no functional redundancy in FUT8 and GMD, although there is a fuctional redundancy in GFT [28,30,33,34]. The loss of GFT seemed to be compensated for by other genes possessing transporter activity. In ordi-nary cell cultures, FUT8 is considered to be the only ratelimiting step enzyme of fucosylation, since mammalian cells retain excess endogenous intracellular GDP-fucose [30,35]. In theory, the level of product fucosylation should be controlled, either in a manner dependent on endogenous FUT8 activity under cellular conditions of abundant intracellular GDP-fucose, or in a manner dependent on endogenous GMD activity under condition of intracellular GDP-fucose starvation; these genes are not currently believed to act in an interdependent manner. However, the present results clearly demonstrated that Serum-free fed-batch culture of clones transformed by the FUT8 and GMD tandem siRNA expression vector  Biological activity of antibodies from clones transformed with the FUT8 and GMD siRNA tandem expression vector Single-step conversion of already-established antibodyproducing cells to non-fucosylated antibody producers by double knockdown of FUT8 and GMD was attempted in order to verify the industrial applicability of such a system. Our transformation strategy was found to function quite well; in brief, we introduced an siRNA tandem expression vector for introducing the double knockdown of FUT8 and GMD into antibody-producing cells, and we carried out the cellular phenotypic selection of resistance to LCA in the presence of L-fucose. We successfully established antibody-producing clones in which no fucosylated oligosaccharides were detected by either monosaccharide composition or MALDI-TOF MS analyses (Table 3, 4, Fig. 3D, 3E). Due to the very high conversion frequency achieved with this approach, only two months were required to complete all steps of the conversion; the ratio of LCA-resistant clones to drug-resistant clones transformed by the siRNA tandem expression vector was higher than that of clones transformed by each single siRNA expression unit (Table 1), and six clones out of the twenty LCA-resistant transformants randomly selected were found to be converted precisely to the desired clones (Table 3). The established converted clones proved capable of consistently producing fully non-fucosylated antibodies throughout the duration of a serumfree fed-batch culture period; furthermore, a constant ratio of F0G0, F0G1, and F0G2 oligosaccharides was maintained until the end of the culture period (Fig. 3, Table 4). The antibody productivity of the clones was also found to be equivalent to that of the parental cells at the end of the culture period. The purified antibodies obtained from the cultures exhibited approximately 100fold higher ADCC compared to that of the original antibodies obtained from the parental cells (Fig. 4). The stability of the converted clones was confirmed for a long culture period of over 30 passages with stable integration of the siRNA expression unit into the genome (data not shown). The production and purification processes established for the 1 st -generation therapeutic antibodies could be applied, with only marginal modifications, to establish the 2 nd -generation processes. Thus, the present approach could provide substantial time and cost benefits for the development of next-generation therapeutic antibodies.
Our preliminary experiments have also indicated that the tandem siRNA expression vector designed for the double knockdown of mouse FUT8 and GMD also worked in mouse myeloma cell lines NS0 and SP2/0.

Conclusion
The double knockdown of FUT8 and GMD in antibodyproducing cells is worth considering as a novel strategy for generating therapeutic antibodies fully lacking core fuco-sylation with enhanced ADCC. The tremendous cost-and time-sparing advantages of this approach are expected to facilitate the development of next-generation therapeutic antibodies. This approach is based on our observation of the synergistic effects of double knockdown of FUT8 and GMD on intracellular oligosaccharide fucose modification in mammalian cells.

Analyses of antibody-derived N-linked Fc oligosaccharides
Recombinant antibodies were purified from the serumfree culture supernatant by Protein A-affinity chromatography using MabSelect™ (Amersham Biosciences, Piscataway, NJ) and the samples were stored in 10 mM KH 2 PO 4 . The concentration of purified antibodies was measured by absorbance at 280 nm. The monosaccharide composition of each purified IgG1 was characterized by modified highperformance anion exchange chromatography (HPAEC) as previously described [8]. The oligosaccharide profile of each purified antibody was characterized by modified matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) in positive-ion mode, as described previously [29].

Biological activity analysis of antibodies
An ADCC assay for each purified IgG1 was performed by lactate dehydrogenase (LDH) release assay using human peripheral blood mononuclear cells (PBMCs) from healthy donors as effector cells at an E:T ratio of 20:1, as described previously [26]. The antigen-binding activities of the antibodies were measured by antigen-binding ELISA, as previously described [26].

Isolation of clones transformed by two siRNA expression vectors
The isolated GMD siRNA-introduced clones (1.6 × 10 6 cells) were electroporated again with 10 μg of FUT8 or GFT siRNA expression vector linearized at the FspI site (FT8lib3/pAGE or GFT_3G10/pAGE). Transfectants were selected in 3 μg/mL puromycin and 400 μg/mL hygromycin (Wako Pure Chemical Industries, Osaka, Japan) for 8 days, and then the drug-resistant clones were subjected to 7-day selection with 500 μg/mL LCA in the presence of 100 μM L-fucose. The resultant LCA-resistant clones were isolated and expanded in IMDM medium containing 10% (v/v) dFBS, 500 nM MTX, 3 μg/mL puromycin, and 400 μg/mL hygromycin for further real-time PCR and antibody analyses, as described above.