A Cancer-specific Monoclonal Antibody Recognizes the Aberrantly Glycosylated Podoplanin

Podoplanin (PDPN/Aggrus/T1α), a platelet aggregation-inducing mucin-like sialoglycoprotein, is highly expressed in many cancers and normal tissues. A neutralizing monoclonal antibody (mAb; NZ-1) can block the association between podoplanin and C-type lectin-like receptor-2 (CLEC-2) and inhibit podoplanin-induced cancer metastasis, but NZ-1 reacts with podoplanin-expressing normal cells such as lymphatic endothelial cells. In this study, we established a cancer-specific mAb (CasMab) against human podoplanin. Aberrantly glycosylated podoplanin including keratan sulfate or aberrant sialylation, which was expressed in LN229 glioblastoma cells, was used as an immunogen. The newly established LpMab-2 mAb recognized both an aberrant O-glycosylation and a Thr55-Leu64 peptide from human podoplanin. Because LpMab-2 reacted with podoplanin-expressing cancer cells but not with normal cells, as shown by flow cytometry and immunohistochemistry, it is an anti-podoplanin CasMab that is expected to be useful for molecular targeting therapy against podoplanin-expressing cancers.

Characterization of the LpMab-2 mAb. Because the binding affinity of antibodies is critical for antibody-based cancer therapy, the dissociation constant (K D ) was next determined using ELISA and flow cytometric analysis 26 . The K D of LpMab-2 was measured as 1.1 3 10 29 M using ELISA and was measured as 5.7 3 10 29 M against LN319 cells and 3.5 3 10 29 M against LN229/hPDPN cells using flow cytometry (Figure 3a, 3b, 3c). The K D values of the other mAbs are also shown in Figure 3a, 3b, and 3c. The binding affinity of LpMab-2 was the best of the four mAbs in the flow cytometric analyses (Figure 3b and 3c), although the affinity of LpMab-2 was worse than those of LpMab-3 and LpMab-7 in ELISA ( Figure 3a). We next performed a kinetic analysis of the interaction of LpMab-2 with a recombinant podoplanin using surface plasmon resonance (BIAcore) 27 . Determination of the association and dissociation rates from the sensorgrams revealed that a k assoc of 3.94 3 10 5 (mol/L-s) 21 and a k diss of 4.82 3 10 23 s 21 . The K A at binding equilibrium, calculated as K A 5 k assoc /k diss , was 8.17 3 10 7 (mol/L) 21 , K D 5 1/K A 5 1.22 3 10 28 M.
The LpMab-2 epitope was determined to be 55-80 amino acids using ELISA (Figure 2a), and sialylated O-glycan was included in the LpMab-2 epitope ( Figure 2b). Therefore, we produced point mutations at the Ser/Thr residues in amino acids 54-88 of podoplanin ( Figure 3d). Of the 12 Ser/Thr residues in this region, the reaction of LpMab-2 against the T55A and S56A mutants was very low, while the other mutants were recognized by LpMab-2, indicating that Thr55 and Ser56 are included in the LpMab-2 epitope. We also produced several point mutations around Thr55 and Ser56. LpMab-2 did not react with the E57A, D58A, R59A, Y60A, or L64A mutants. Taken together, the evidence indicates that the important podoplanin epitope for LpMab-2 is the glycopeptide Thr55-Leu64: TSEDRYKTGL (Figure 3e; right). The sequencing gap was observed at several Ser/Thr residues during Edman degradation of podoplanin purified from CHO/hPDPN cells, and wellglycosylated Ser/Thr residues were identified (Figure 3e; left) 21 . The glycosylation of Thr55 and Ser56 was not detected in CHO/hPDPN cells, indicating that only a small population of human podoplanin is glycosylated at Thr55 and Ser56 (Figure 3e; right).
We previously demonstrated that the NZ-1 mAb can be internalized into LN319 glioblastoma cells 27 . Similarly, LpMab-2 was internalized into podoplanin-positive LN319 cells but not into podoplanin-negative LN229 cells (Figure 3f), indicating that LpMab-2 is a candidate for use in an antibody-drug conjugate.
Immunohistochemical analyses of LpMab-2 against podoplaninexpressing cancers and normal tissues. Because LpMab-2 was www.nature.com/scientificreports SCIENTIFIC REPORTS | 4 : 5924 | DOI: 10.1038/srep05924   After blocking with SuperBlock T20 (PBS) Blocking Buffer, the plates were incubated with LpMab-2, LpMab-3, LpMab-7, and LpMab-9 (1 mg/ml) followed by 1:1,000 diluted peroxidase-conjugated anti-mouse IgG. The enzymatic reaction was conducted with 1-Step Ultra TMB-ELISA. The optical density was measured at 655 nm using an iMark microplate reader. These reactions were performed in a volume of 50 ml at 37uC. The error bars show the standard deviation of three independent experiments. (b) Characterization of anti-podoplanin mAbs. Cell lysates (10 mg) were boiled in SDS sample buffer, then electrophoresed on 5-20% polyacrylamide gels and transferred onto a PVDF membrane. After blocking, the membrane was incubated with primary antibodies (1 mg/ml and then with peroxidase-conjugated secondary antibodies; the membrane was developed with ECL-plus reagents using a Sayaca-Imager. Arrows indicate a 35-kDa band of podoplanin. (c, d) Flow cytometry using anti-podoplanin mAbs against cancer cells (c) and normal cells (d). Cell lines were harvested by brief exposure to 0.25% Trypsin/1 mM EDTA. After washing with PBS, the cells were treated with primary antibodies (1 mg/ml) for 30 min at 4uC followed by treatment with Oregon green-conjugated anti-mouse IgG. Fluorescence data were collected using a cell analyzer (EC800; Sony Corp.  human IgG 1 -Fc) was immobilized at 1 mg/ml. The plates were incubated with serially diluted LpMab-2, LpMab-3, LpMab-7, and LpMab-9 (150 pg/ml-2.5 mg/ml) followed by 1:1,000 diluted peroxidase-conjugated anti-mouse IgG. The dissociation constants (K D ) were obtained by fitting the binding isotherms using the built-in one-site binding models in Prism software. (b, c) Determination of binding affinity using flow cytometry. LN319 or LN229/hPDPN cells (2 3 10 5 cells) were resuspended with 100 ml of serially diluted LpMab-2, LpMab-3, LpMab-7, and LpMab-9 (0.02-100 mg/ml) followed by secondary anti-mouse IgG. Fluorescence data were collected using an EC800 Cell Analyzer. The dissociation constants (K D ) were obtained by fitting the binding isotherms using the built-in one-site binding models in Prism software. (d) Epitope mapping of LpMab-2 using flow cytometric analyses. Podoplanin mutant-expressing CHO cells were harvested by brief exposure to 0.25% Trypsin/1 mM EDTA. After washing with PBS, cells were treated with LpMab-2 and LpMab-7 (1 mg/ml) for 30 min at 4uC, followed by treatment with Oregon green-conjugated anti-mouse IgG. Fluorescence data were collected using an EC800 Cell Analyzer. (e) Left: The sequencing gap was observed at several Ser/Thr residues in human podoplanin purified from CHO/hPDPN cells by Edman degradation. Well-glycosylated Ser/Thr residues (Thr52, Thr65, Thr66, Thr70, Ser71, Ser74, Thr76, Ser98, Thr100, Ser107, Ser109, and Thr110) were identified. PLAG domains and O-glycans are shown in yellow and blue, respectively. Right: Schematic illustration showing dual recognition of LpMab-2 against podoplanin glycopeptide. (f) Internalization assay of LpMab-2 with podoplanin positive-LN319 cells and podoplanin negative-LN229 cells. LN319 and LN229 glioblastoma cells were plated on a 24-well plate and were incubated for 24 h. Then, LpMab-2-pHrodo was added to the medium (30 mg/ml). The cells were incubated for 50 h, then were washed once with PBS. Fluorescence microscopy was performed using a FLoid Cell Imaging Station. The nucleus was stained with DAPI. determined to be a CasMab by flow cytometry, we performed immunohistochemical analyses of established anti-podoplanin mAbs. LpMab-2 reacted with cancer cells, not with lymphatic endothelial cells in esophageal squamous cell carcinomas (Figure 4a) and seminoma tissues (Figure 4b). By contrast, LpMab-7 stained both cancer cells and lymphatic endothelial cells in esophageal squamous cell carcinomas (Figure 4f) and seminoma tissues (Figure 4g). Both LpMab-2 and LpMab-7 stained glioblastoma cells (Figure 4c and 4h). Immunostaining of LpMab-2 and LpMab-7 against podoplanin demonstrated predominantly cell-surface patterns in cancer cells (Figure 4a, 4c, 4f, 4h). Proliferating endothelial cells were negative for podoplanin (Figure 4c and 4h). LpMab-7 reacted with normal esophageal lymphatic endothelial cells (Figure 4i) and lung type I alveolar cells (Figure 4j), whereas LpMab-2 did not (Figure 4d and 4e), indicating that LpMab-2 also acts as a CasMab in immunohistochemistry. LpMab-3 is also useful for immunohistochemical analyses (data not shown). Not all cancer cells are necessarily aberrantly glycosylated; therefore, the intensity of LpMab-2-staining (Figure 4a, 4b, and 4c) was weaker than LpMab-7 (Figure 4f, 4g,  and 4h).

Discussion
As we previously reported, an anti-podoplanin mAb, NZ-1, was highly internalized into glioma cell lines and also accumulated efficiently into tumors in vivo 27 . NZ-1 and its rat-human chimeric antipodoplanin antibody (NZ-8) possess antibody-dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC) functions against podoplanin-expressing glioblastoma or malignant mesothelioma cell lines 16,26 . Furthermore, NZ-1 inhibited tumor cell-induced platelet aggregation and tumor metastasis by its neutralizing activity 13 , indicating that NZ-1 is a suitable candidate for molecular targeted therapy against podoplanin-expressing cancers. Although many anti-podoplanin mAbs have been produced, those mAbs, including NZ-1, react with podoplanin-expressing normal cells, including lung type I alveolar cells, renal podocytes, mesothelial cells, and lymphatic endothelial cells throughout the body 13,22,23,[28][29][30] . Recently, critical physiological functions of podoplanin have been reported [3][4][5]8 . Therefore, a cancer-specific anti-podoplanin mAb is necessary for molecular targeted therapy against podoplaninexpressing cancers. In our previous study, we revealed that LN229 cells possess cancer-type glycosylation patterns, including highly sulfated keratan sulfate (KS), which are similar to glioblastoma tissues 31,32 . Highly sulfated KS, which possesses a polylactosamine structure, was detected in glioblastoma tissues but not in normal brain or low-grade glioma tissues 32 . Therefore, we used a cancer-type podoplanin, which is expressed in LN229/hPDPN cells. Highly sulfated KS may lead to conformational changes in podoplanin because of negative charges; therefore, immunization with highly sulfated KS-possessing podoplanin is a critical method to induce cancerspecific mAbs against podoplanin. Furthermore, aberrant glycosylation, which is observed in LN229/hPDPN cells, may include not only highly sulfated KS but also other several types of glycans, such as aberrant O-glycosylation and aberrant sialylation. Although highly sulfated KS was detected by several lectins, such as LEL, STL, and UDA, aberrant O-glycosylation or aberrant sialylation were not detected by any lectins using lectin microarray or lectin-blot analyses if they were only attached at different sites with normal type cells and were not over-glycosylated. As we reported previously, Thr55 and Ser56 were non-glycosylated or minimally glycosylated in human podoplanin purified from CHO/hPDPN cells 21 . In addition, LpMab-2 recognized CHO/hPDPN cells at a lower intensity than LN229/hPDPN cells (Figure 2b), indicating that LpMab-2 reacted with cancer-type aberrant glycosylation (O-glycosylation or sialylation, not keratan sulfate) of Thr55 and/or Ser56, which is wellglycosylated in LN229/hPDPN cells and partially glycosylated in CHO/hPDPN cells (Figure 3e). Because LpMab-2 reacted with can-  cer cells and not with lymphatic endothelial cells and type I alveolar cells, LpMab-2 is a cancer-specific mAb. Taken together, the results show that LpMab-2 is expected to be useful for molecular targeting therapy against podoplanin-expressing cancers. We recently produced a single chain Fv fragment (scFv) of an anti-podoplanin mAb (NZ-1) conjugated with immunotoxin, which had an antitumor effect in a glioblastoma xenograft model 33 . That report revealed that anti-podoplanin mAbs are effective against glioblastoma after crossing the blood-brain barrier. Likewise, we will perform further experiments using the scFv-immunotoxin of LpMab-2 to investigate its anti-tumor effects against glioblastoma. Lectin microarray. Podoplanins from LN229/hPDPN and CHO/hPDPN cells were solubilized using 1% Triton-X100 in PBS (PBST) and were purified using a FLAG-tag system (Sigma-Aldrich Corp., St. Louis, MO). Podoplanin from LN319 cells was purified using NZ-1 mAb 21 . Then, 100 ml of purified podoplanin (31.25-2,000 ng/ ml) was applied to a lectin array (LecChip ver1.0; GlycoTechnica, Hokkaido, Japan), including triplicate spots of 45 lectins in each of seven divided incubation baths on the glass slide 19 . After incubation at 20uC for 17 h, the reaction solution was discarded. The glass slide was scanned using a GlycoStation Reader 1200 (GlycoTechnica) 13  Western-blot analyses. Cell lysates (10 mg) or purified podoplanin (0.1 mg) were boiled in SDS sample buffer (Nacalai Tesque, Inc., Kyoto, Japan) 26 . The proteins were electrophoresed on 5-20% polyacrylamide gels (Wako Pure Chemical Industries Ltd.) and were transferred onto a PVDF membrane (EMD Millipore Corp., Billerica, MA). After blocking with SuperBlock T20 (PBS) Blocking Buffer (Thermo Fisher Scientific Inc.), the membrane was incubated with primary antibodies or biotinylated lectin (1 mg/ml; Vector Laboratories Inc., Peterborough, UK), then with peroxidaseconjugated secondary antibodies (Dako; 1/1,000 diluted) or streptavidin-HRP (Dako; 1/1,000 diluted), and developed with the ECL-plus reagent (Thermo Fisher Scientific Inc.) using a Sayaca-Imager (DRC Co. Ltd., Tokyo, Japan).
Production of podoplanin mutants. The amplified human PDPN cDNA was subcloned into a pcDNA3 vector (Life Technologies Corp.) and a FLAG epitope tag was added at the C-terminus. Substitution of amino acids to alanine in podoplanin was performed using a QuikChange Lightning site-directed mutagenesis kit (Agilent Technologies Inc., Santa Clara, CA) 34 . CHO cells were transfected with the plasmids using a Gene Pulser Xcell electroporation system (Bio-Rad Laboratories Inc.).