The Influence of Artificially Introduced N-Glycosylation Sites on the In Vitro Activity of Xenopus laevis Erythropoietin

Erythropoietin (EPO), the primary regulator of erythropoiesis, is a heavily glycosylated protein found in humans and several other mammals. Intriguingly, we have previously found that EPO in Xenopus laevis (xlEPO) has no N-glycosylation sites, and cross-reacts with the human EPO (huEPO) receptor despite low homology with huEPO. In this study, we introduced N-glycosylation sites into wild-type xlEPO at the positions homologous to those in huEPO, and tested whether the glycosylated mutein retained its biological activity. Seven xlEPO muteins, containing 1–3 additional N-linked carbohydrates at positions 24, 38, and/or 83, were expressed in COS-1 cells. The muteins exhibited lower secretion efficiency, higher hydrophilicity, and stronger acidic properties than the wild type. All muteins stimulated the proliferation of both cell lines, xlEPO receptor-expressing xlEPOR-FDC/P2 cells and huEPO receptor-expressing UT-7/EPO cells, in a dose-dependent manner. Thus, the muteins retained their in vitro biological activities. The maximum effect on xlEPOR-FDC/P2 proliferation was decreased by the addition of N-linked carbohydrates, but that on UT-7/EPO proliferation was not changed, indicating that the muteins act as partial agonists to the xlEPO receptor, and near-full agonists to the huEPO receptor. Hence, the EPO-EPOR binding site in X. laevis locates the distal region of artificially introduced three N-glycosylation sites, demonstrating that the vital conformation to exert biological activity is conserved between humans and X. laevis, despite the low similarity in primary structures of EPO and EPOR.


Introduction
Erythropoietin (EPO) is a hematopoietic cytokine that regulates the rate of red blood cell production [1,2]. It binds to EPO receptors (EPORs) on the surface of erythroid progenitors, and Asn24 Thr26, Asp38 Met40 to Asn38 Thr40, and Asp83 to Asn83 (Fig 1A and Table 1). N-glycosylated xlEPO cDNAs were produced by in vitro multiple site-directed mutagenesis of the xlEPO cDNA, as described below.

Multiple site-directed mutagenesis and preparation of expression plasmids
Mutagenesis was performed as described by Seyfang and Jin [38], with some modifications. Double-stranded plasmid DNA containing the xlEPO gene [21] was used as a template. Mutagenic primers and tailed 5 0 -and 3 0 -anchor primers (Table 1) were phosphorylated before the annealing reaction. Two terminal-tailed primers with a unique 17-nucleotide tail (Table 1) were simultaneously annealed to the template DNA along with the set of mutagenic primers. Mutant strands were synthesized by primer extension and ligation in a single incubation step using T4 DNA polymerase and ligase. Full-length mutant strands were subsequently amplified by high-fidelity PCR with Pfu polymerase, which uses the unique mutant strand-specific tails introduced by the 2 terminal anchor primers. The specificity and band size of the amplified mutagenesis products were confirmed by 1.5% agarose gel electrophoresis. Following gel extraction and purification (QIAEX II Gel Extraction Kit; QIAGEN K.K., Tokyo, Japan) of the 584-bp band, the extracted DNA was cloned into a pGEM-T Easy vector (Promega K.K., Tokyo, Japan) and sequenced using an ABI3100 Genetic Analyzer (Applied Biosystems) and dye-terminator chemistry (BigDye Terminator; Applied Biosystems). Plasmids containing the desired mutations were digested with NheI and EcoRI. The DNA fragments were then cloned into the multiple cloning site (MCS) of the pIRES2-EGFP expression vector (BD Biosciences Clontech) containing the human cytomegalovirus immediate early promoter and an SV40 early polyadenylation signal sequence. pIRES2-EGFP contains the internal ribosome entry site (IRES) of the encephalomyocarditis virus between the MCS and the EGFP coding region; this enables both the xlEPO gene and the EGFP gene to be translated from a single bicistronic mRNA.
All constructs were transformed into E. coli DH5α. Several colonies were subsequently grown in Luria-Bertani broth containing 50 μg/mL of kanamycin. The plasmids were isolated using the alkaline-SDS lysis method.

Expression of xlEPO muteins in COS-1 cells
Monkey kidney fibroblast COS-1 cells (Cell Bank, RIKEN BioResource Center, Ibaraki, Japan) were transfected with vectors that incorporated the full coding sequence of the xlEPO genes. COS-1 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% (v/v) FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin in a humidified atmosphere of 5% CO 2 at 37°C. Cells (8 × 10 4 cells/well) were seeded in 24-well cell culture plates (Corning Incorporated, NY, USA). When the cultures reached 90% confluence, cells were transfected with a mixture of 0.8 μg plasmid cDNA and 1.5 μL Lipofectamine LTX (Life Technologies Corporation) per well in 100 μL Opti-MEM (Life Technologies Corporation) for 4 h at 37°C. After transfection, cells were incubated in antibiotic-free DMEM containing 10% FBS for 24 h. Cells were then incubated in serum-free, antibiotic-containing DMEM. After an additional 48 h, EGFP fluorescence was measured (excitation 485 nm, emission 528 nm) and conditioned media were collected. The collected media were arbitrarily concentrated using centrifugal filter devices (Amicon Ultra-4, 5000 NMWL; Millipore Corporation, Billerica, MA, USA) and stored at-80°C. Expression of the xlEPO muteins was confirmed by SDS-PAGE followed by Western blotting.  [31] or anti-GFP antibody (Medical & Biological Laboratories Co., Ltd., Nagoya, Japan) and then washed twice with TTBS. The blots were subsequently incubated for 1 h with biotinylated goat anti-rabbit IgG (1:6000; Bio-Rad), washed twice with TTBS, and treated with alkaline phosphatase-conjugated StrepTactin (1:6000; Bio-Rad) for 1 h. After 2 washes with TTBS followed by 2 more washes with TBS, proteins were detected by chemiluminescence (CDP-Star reagent; GE Healthcare Ltd., Tokyo, Japan). The bands were detected using a Luminescent Image analyzer LAS-3000 (Fuji Photo Film Co., Ltd., Tokyo, Japan). All washes were carried out for 5 min. All antibodies and alkaline phosphatase-conjugated StrepTactin were prepared in TTBS containing 0.4% Block Ace.

Physicochemical analysis of xlEPO muteins
Wild-type xlEPO and N-glycosylated muteins in mixed conditioned media were separated by reverse-phase high-pressure liquid chromatography (RP-HPLC) and cation exchange chromatography (CIEC). In RP-HPLC, the sample was applied at a flow rate of 0.4 mL/min to a YMC-Pack Protein-RP column (4.5 × 150 mm; YMC Co., Ltd., Tokyo, Japan) and separated with a linear gradient of 20%-60% acetonitrile in 0.1% trifluoroacetic acid over 30 min. A total of 10 fractions was collected once every 3 min (1.2 mL). In CIEC, the samples were adjusted to pH 2 with acetic acid, applied to 200 μL of SP Sepharose FF column (GE Healthcare), and eluted with 50 mM citrate/100 mM phosphate buffer (pH 2, 3, 4. . . 9). Each fraction was subjected to Western blot analysis.
De-N-glycosylation of xlEPO muteins N-linked carbohydrates were removed by incubation with or without 0.8 U peptide N-glycosidase F (PNGase F; Roche Diagnostics K.K., Tokyo, Japan) in 20 mM sodium phosphate buffer (pH 8.6) containing 10 mM EDTA at 37°C for 24 h. Carbohydrate removal was assessed by SDS-PAGE and Western blot analysis.

Cell proliferation assay
Cell proliferation assays were performed as previously described [21]. Two cell lines were used: the interleukin-3-dependent murine cell line FDC/P2, which expresses the exogenous xlEPO receptor, xlEPOR-FDC/P2 [21], and the EPO-dependent human cell line, UT-7/EPO [39]. In the MTS assay, absorbance at 450 nm was measured using a microplate reader (POWERSCAN HT; DS Pharma Biomedical Co., Ltd., Osaka, Japan). Dose-response curves were fitted to a 4-parameter logistic model: where, a is E max and c is EC 50 [40]. Then, the E max (maximum effect) of the xlEPO mutein was determined without normalization by protein amount. The relative EC 50 (half-maximal effective concentration) of the xlEPO mutein was determined via normalization of v/v % of COS-1 supernatant EC 50 with its relative amount in COS-1 supernatant.

Statistical analysis
All experiments were performed at least 3 times with triplicate samples, and similar results were obtained each time. Error bars represent the standard deviations (SD). Either Student's ttest or the Tukey-Kramer test was used, when appropriate.

Design and expression of N-glycosylated xlEPO muteins
We introduced N-glycosylation consensus sequences into the xlEPO at positions homologous to those of huEPO (i.e., Asn24, Asn38, and Asn83) using in vitro mutagenesis ( Fig 1A). The mutations were introduced both individually and in combinations. Thus, we constructed 7 different N-glycosylated xlEPO muteins with consensus sequences introduced at 3 different positions ( Fig 1B). The wild-type xlEPO and N-glycosylated muteins were expressed in COS-1 cells using the lipofection method. All muteins were secreted from the COS-1 cells. The extent of glycosylation was evaluated by Western blot analysis using the anti-xlEPO sequence antibody (Fig 2A, top panel). The molecular weight of wild-type xlEPO was 18 kDa, which is equivalent to that of the peptide backbone. Western blots showed progressive increases in the molecular weights of xlEPO muteins from 18 to 22, 26, and 30 kDa, depending on the number of N-linked carbohydrates added. Treatment with PNGase F, which removes N-linked carbohydrates from glycoproteins, shifted the molecular weights of all muteins to 18 kDa, or the molecular weight of the wild type (Fig 2A, bottom panel). Thus, the increase in the size of the muteins in comparison to the wild-type xlEPO was attributed to N-linked carbohydrate content. Non-glycosylated Asn83 was observed in xlEPO-3, xlEPO-13, xlEPO-23, and xlEPO-123, which indicates that carbohydrate additions were efficient at Asn24 and Asn38, but inefficient at Asn83. The glycosylated xlEPO molecules showed 2 bands: a main band and a smaller one. These 2 bands were not observed after PNGase F treatment. This result shows that these 2 bands result from heterogeneity in the N-linked carbohydrate structure. The unsecreted protein was also examined by Western blot analysis ( Fig 2B). Whole-cell lysates were loaded into each lane, and EGFP was used as a loading control to normalize the signals (Fig 2B, bottom panel). The wild-type xlEPO and N-glycosylated muteins left in the cell were detected (Fig 2B, top panel). To estimate the amount of xlEPO muteins in the COS-1 supernatant, we performed a densitometry analysis of Western blots. In Western blot analysis, the recognition of hyperglycosylated huEPO analogs by an anti-huEPO sequence antibody was reported to be weak compared with huEPO recognition, and was improved by the partial elimination of carbohydrate chains [41]. Therefore, we used Western blots of deglycosylated samples (Fig 2A, bottom panel) for densitometry analysis, because the added N-linked carbohydrates may prevent the anti-xlEPO sequence antibody from binding to the xlEPO muteins. The band intensity of wild-type xlEPO was defined as 1.
The analysis revealed that the relative amount of xlEPO muteins in the COS-1 supernatant decreased depending on the number of N-linked carbohydrates added (Fig 2C). The detected amounts of the muteins, xlEPO-1, xlEPO-2, and xlEPO-3, containing 1 N-glycosylation site were approximately half that of the wild type. Moreover, the amounts of the muteins, xlEPO-12, xlEPO-23, and xlEPO-123, containing 2 or 3 N-glycosylation sites were only about one-quarter that of the wild type. Because the IRES-containing expression vector enables xlEPO and EGFP to be co-expressed from a single bicistronic mRNA, EGFP fluorescence intensity reflects xlEPO expression at the mRNA level. Therefore, the relative secretion efficiency of xlEPO muteins was obtained by dividing the relative amount (i.e. relative secretion level) by the relative expression level (Fig 2C). As a result, the secretion efficiencies of xlEPO muteins were revealed to decrease as the number of N-linked carbohydrates added increased, suggesting that glycosylation may have an adverse effect on xlEPO secretion. Physicochemical properties of xlEPO muteins The physicochemical properties of proteins, such as the electrical charge and hydrophilicity, affect their ligand-receptor binding affinity and/or protein stability in blood [42,43]. To predict the isoelectric point (pI) of the xlEPO muteins, CIEC was performed (Fig 2D, top panel). Wild-type xlEPO was eluted at pH 8-9, and the xlEPO muteins containing 1 or 2 N-linked carbohydrates were eluted mainly at pH 7 or 8, respectively. Next, RP-HLPC was performed, in order to examine the hydrophilicity of the xlEPO muteins (Fig 2D, bottom panel). Wildtype xlEPO was eluted in fraction (Fr.) 9 at 52%-58% acetonitrile. Meanwhile, xlEPO muteins with 1 or 2 N-linked carbohydrates were eluted in earlier fractions (Fr. 8, 46%-52% acetonitrile or Fr. 7, 40%-46% acetonitrile, respectively). These results indicate that the xlEPO muteins are more acidic and hydrophilic than the wild type, owing to the negative charge of N-linked carbohydrates.

In vitro activity of xlEPO muteins
The effects of the added N-linked carbohydrates on the in vitro activity of xlEPO muteins were tested in cell proliferation assays using exogenous xlEPOR-expressing xlEPOR-FDC/P2 cells (Fig 3A), and huEPOR-expressing UT-7/EPO cells ( Fig 3B). As expected, all xlEPO muteins stimulated the proliferation of both cell lines in a dose-dependent manner. Usually, dose-response curves as descriptors of ligand activity are characterized by differences in potency (described as EC 50 ) and efficacy (described as E max ). Then, the relative potency and the relative efficacy of the xlEPO muteins on the proliferation of xlEPOR-FDC/P2 and UT-7/EPO cells were evaluated (Table 2 and Fig 4). As the number of N-linked carbohydrates increased, the relative potency of xlEPO muteins, which was obtained from the relative EC 50 , tended to decrease in both cell lines (Fig 4A and 4B). On the other hand, the effect of N-linked carbohydrates on the relative efficacy of xlEPO muteins, which was obtained from E max , differed between the cell lines (Fig 4C and 4D). The efficacy on the proliferation of xlEPOR-FDC/P2 cells tended to decrease, as the number of N-linked carbohydrates increased (Fig 4C). More specifically, the efficacy of xlEPO-12 and xlEPO-23 was significantly lower than that of wildtype xlEPO, and xlEPO-1 and xlEPO-13 also showed a slightly lower efficacy compared to the wild type (Table 2). In contrast, the efficacy of the xlEPO muteins on the proliferation of UT-7/ EPO cells was not significantly affected by the addition of N-linked carbohydrates (Fig 4D and  Table 2). To confirm that the decrease in the activity of xlEPO muteins (EC 50 and E max ) is attributed to the addition of N-linked carbohydrates, PNGase F-treated COS-1 supernatant was assayed by the proliferation of xlEPOR-FDC/P2 cells (Fig 3C). Removal of N-linked carbohydrates with PNGase F increased the in vitro activity of the xlEPO muteins (Fig 3C, middle and right panels), except xlEPO-3 (Fig 3C, left panel), when COS-1 supernatants were added at a final concentration of 6%. This indicates that the added N-linked carbohydrates decreased the in vitro activity of xlEPO.

Discussion
In this study, we produced 7 different N-glycosylated xlEPO muteins and compared their biochemical properties of (i) secretion, (ii) physicochemical properties, and (iii) in vitro activity to those of the wild type.
The relative secretion efficiencies of the xlEPO muteins decreased depending on the number of N-linked carbohydrates added (Fig 2C). When the carbohydrate is added to the partially folded protein during translation [44], it has been hypothesized that the rates of synthesis or folding have a competitive relationship with the addition of the carbohydrate [45]. In light of this, the addition of carbohydrates was thought to reduce the secretion of xlEPO muteins at translation and/or during the secretion process.
Wild-type xlEPO was absorbed to the cation exchanger at pH 7 ( Fig 2D). The pH of X. laevis plasma is approximately 7.4, the same as that of human plasma [46]. This indicates that wildtype xlEPO bears a positive electric charge in blood, whereas huEPO (pI, 4.4-5.1) bears a negative charge [47]. In case of huEPO, sialic acid of carbohydrates, which is negatively charged at physiological pH, is thought to decrease EPOR binding via charge repulsion. Accordingly, wild-type xlEPO is thought to easily access EPORs on cells compared to huEPO.
All N-glycosylated xlEPO muteins retained in vitro activity against the cell proliferation of both xlEPOR-FDC/P2 and UT-7/EPO (Fig 3A and 3B). The EC 50 and E max of a ligand represent the ability of the ligand to bind to its receptor (i.e., potency which reflects receptorbinding affinity) and its ability to cause an effect after binding to the receptor (i.e., efficacy which is intrinsic acticity), respectively. In the present study, we showed the relative EC 50 because we were able to roughly quantify the relative amounts of wild-type xlEPO and N-glycosylated muteins in conditioned media. As a result, we found that the addition of N-linked carbohydrates decreased the EC 50 -derived relative potency of xlEPO muteins (Fig 4A and 4B). It has been reported that N-linked carbohydrates reduce the affinity of huEPO for its receptor, as well as the in vitro activity of huEPO [5]. This can be explained as follows: the negatively charged sialic acid and carbohydrate individually decrease the association rate constant of huEPO with huEPOR via their negative effect on intrinsic electrostatic enhancement [42]. It is also known that darbepoetin alfa, a hyperglycosylated huEPO analog with 2 extra N-linked carbohydrates, shows lower receptor-binding activity than huEPO [48]. Here, we confirmed that the N-glycosylated xlEPO muteins are more negatively charged than the wild type ( Fig 2D). Taken together, it is thought that one of the reasons for reduced potencies of muteins is their lower receptor-binding affinities than that of wild type due to negative charge of carbohydrates.
Type of ligand effect on a receptor can be classified according to its behavior. Full agonists elicit the same level of full effect (i.e. E max , efficacy) as the natural ligand of the receptor, while partial agonists also cause an effect, but they cannot reach the same level of the natural ligand. The efficacy of the xlEPO muteins on xlEPOR-FDC/P2 cell proliferation was less than that of wild-type xlEPO (Fig 4C). In particularly, the efficacy of the xlEPO muteins containing Nlinked carbohydrates on Asn24 was remarkably lower than that of wild-type xlEPO ( Table 2). These data indicate that the muteins act as partial agonists against xlEPOR, and also suggest that the carbohydrate structure located at Asn24 interferes with xlEPOR binding. In contrast, the efficacy of xlEPO on the proliferation of UT-7/EPO cells was not changed by the addition of N-linked carbohydrates (Fig 4D). The E max of wild-type xlEPO and the N-glycosylated muteins was approximately the same as that of huEPO. These data indicate that the muteins act as near-full agonists against huEPOR, and also suggest that the N-linked carbohydrates at the Asn24, Asn38, and Asn83 sites of the xlEPO muteins do not interfere with huEPOR binding. Analysis of the effect of point mutations on bioactivity [26][27][28][29] and the crystal structure of huEPO, complexed with the extracellular ligand-binding domains of huEPOR [30], have identified amino acid residues important for receptor binding and biological activity. Twelve of the 18 amino acid residues that reduce the in vitro bioactivity of huEPO following point mutations, and interact with huEPOR residues, are conserved in xlEPO (6 of 10 in site 1, 6 of 8 in site 2), despite an only 38% overall sequence identity between xlEPO and huEPO (Table 3) [21]. These highly conserved amino acid residues are thought to allow xlEPO to cross-react with huEPOR and retain its proper orientation, avoiding interference from the added N-linked carbohydrates. With regard to xlEPOR, 7 of the 13 amino acid residues of huE-POR that interact with huEPO residues are conserved in xlEPOR (6 of 8 in site 1, 2 of 8 in site 2), which has an only 26% sequence identity with huEPOR in the extracellular domain (Table 3) [49]. Further investigation is needed to understand why the added carbohydrates on xlEPO interfere with the xlEPOR binding but not with the huEPOR binding. The most important finding presented here is that the carbohydrate structures, which are added to those xlEPO sites homologous to the N-glycosylation sites of huEPO, do not completely eliminate the in vitro biological activity of xlEPO, namely its binding to EPOR. This suggests that the functional region of xlEPO, which is responsible for its in vitro activity, must be located in a region other than the introduced N-glycosylation sites, as the 3 N-glycosylation sites of huEPO are distal to the receptor-binding site. We therefore conclude that a fundamental conformation of EPO-EPOR binding could be conserved between X. laevis and humans. Further point mutation study at the xlEPO-xlEPOR/huEPOR interface will give definite conclusions about their active sites. Furthemore, the key to understanding the observations presented here, as well as this conclusion, is the elucidation of the crystal structure of xlEPO in a complex with the extracellular domains of xlEPOR. Cross-species comparison of the tertiary Table 3. Amino acid residues of huEPO and huEPOR that contact each other in site 1 and site 2 intermolecular contact areas, and the corresponding residues in xlEPO. a xlEPO residues corresponding to right-hand "huEPO residues" [21]. Amino acid residues that are identical to huEPO residues are in bold.
c HuEPOR residues within 4.5-Å distance from huEPO in the crystal structure of human EPO-(EPOR) 2 complex (Protein Data Bank accession number, 1EER) [30]. Amino acid residues that are conserved in X. laevis are in bold [49]. d Amino acid residues of huEPO and huEPOR that form interactions with each other, such as hydrogen bond and hydrophobic binding, are in italics [30].
e Amino acid residues of xlEPOR that conserved in huEPOR [49]. structures of X. laevis and human EPOs (and other species such as fish) will facilitate the identification of the common denominators present in vertebrate EPOs.