Anti-podocalyxin antibody exerts antitumor effects via antibody-dependent cellular cytotoxicity in mouse xenograft models of oral squamous cell carcinoma

Podocalyxin (PODXL) overexpression is associated with progression, metastasis, and poor outcomes in cancers. We recently produced the novel anti-PODXL monoclonal antibody (mAb) PcMab-47 (IgG1, kappa). Herein, we engineered PcMab-47 into 47-mG2a, a mouse IgG2a-type mAb, to add antibody-dependent cellular cytotoxicity (ADCC). We further developed 47-mG2a-f, a core fucose-deficient type of 47-mG2a to augment its ADCC. Immunohistochemical analysis of oral cancer tissues using PcMab-47 and 47-mG2a revealed that the latter stained oral squamous cell carcinoma (OSCC) cells in a cytoplasmic pattern at a much lower concentration. PcMab-47 and 47-mG2a detected PODXL in 163/201 (81.1%) and in 197/201 (98.0%) OSCC samples, respectively. 47-mG2a-f also detected PODXL in OSCCs at a similar frequency as 47-mG2a. In vitro analysis revealed that both 47-mG2a and 47-mG2a-f exhibited strong complement-dependent cytotoxicity (CDC) against CHO/hPODXL cells. In contrast, 47-mG2a-f exhibited much stronger ADCC than 47-mG2a against OSCC cells, indicating that ADCC and CDC of those anti-PODXL mAbs depend on target cells. In vivo analysis revealed that both 47-mG2a and 47-mG2a-f exerted antitumor activity in CHO/hPODXL xenograft models at a dose of 100 μg or 500 μg/mouse/week administered twice. 47-mG2a-f, but not 47-mG2a, exerted antitumor activity in SAS and HSC-2 xenograft models at a dose of 100 μg/mouse/week administered three times. Although both 47-mG2a and 47-mG2a-f exerted antitumor activity in HSC-2 xenograft models at a dose of 500 μg/mouse/week administered twice, 47-mG2a-f also showed higher antitumor activity than 47-mG2a. These results suggested that a core fucose-deficient anti-PODXL mAb could be useful for antibody-based therapy against PODXL-expressing OSCCs.


INTRODUCTION
In total, 300,000 cases of oral cancer are reported annually, constituting approximately 2% of all cancer cases globally [1], and several histological tumor types exist, including squamous cell carcinoma (OSCC), adenoid cystic carcinoma, mucoepidermoid carcinoma, and osteosarcoma. Among them, almost 90% of oral cancers are OSCCs [2]. The most common location of OSCC is the tongue, accounting for approximately 40% of OSCCs [3,4]. Despite improvements in diagnostic technology and therapeutic techniques, the survival rate of OSCC has improved by only 5% over the past 20 years. Consequently, the 5-year survival rate is 60% [5], and the incidence of OSCC has increased globally [4,6].
OSCC is mainly treated via surgical removal, which can be complemented by radiotherapy and/or chemotherapy, especially in advanced stages. Many types of anticancer agents, including cisplatin (CDDP), 5-fluorouracil (5-FU), and docetaxel, are used for chemotherapy [7,8]. In contrast, the availability of approved molecular targeting drugs with efficacy against OSCC is limited. Recently, cetuximab, a mouse-human (IgG 1 ) chimeric antibody against epidermal growth factor receptor (EGFR), was approved for treating head and neck cancer (HNC) including oral cancer. In several clinical studies, cetuximab was found to be effective against locoregionally advanced head and neck squamous cell carcinoma (HNSCC) or recurrent and/or metastatic (R/M) HNSCC [7,[9][10][11]. Furthermore, nivolumab, a fully human IgG 4 mAb against programmed cell death-1, was also approved for the treatment of R/M HNC, which has been treated with platinum-based chemotherapy [12]. In addition, bevacizumab, which is a mouse-human IgG 1 chimeric antibody against vascular endothelial growth factor (VEGF) and first approved for colorectal cancer, had some clinical trial in which R/M HNSCC patients were enrolled [13]. Because molecular targeting drugs that are clinically applicable for oral cancers are limited, novel drugs with greater efficacy and lower toxicity are required. Therefore, we have been investigating several membrane proteins, and we developed several mAbs against those targets, including HER2 [14], EGFR [15], podoplanin [16], and PODXL [17].
Despite the development of anti-PODXL monoclonal antibodies (mAbs) [32,33], the efficacy of these treatments against oral cancers remains to be fully elucidated. We previously immunized mice with recombinant PODXL, which was purified from the culture supernatant of LN229/ ectodomain-PODXL cells [17]. One clone, PcMab-47 (mouse IgG 1 , kappa), was successfully produced. We further produced chPcMab-47 from PcMab-47 and investigated its antitumor activity against colorectal cancers [34]. In those studies, we injected human NK cells around the tumors to investigate antitumor activity because chPcMab-47 could not induce ADCC activity using mouse NK cells. Although chPcMab-47 significantly reduced tumor development compared with the effects of control human IgG, its antitumor activity might not be sufficient for antibodybased target therapy. We used human NK cells from one donor; therefore, the cause of the low antitumor activity is due to differences in the sources of cells. Because we had to inject human NK cells around the subcutaneous tumors of xenograft models several times, we experienced difficulties in measuring tumor diameters because the tumor shape sometimes changed post-injection. Furthermore, mouse IgG 1 does not induce ADCC and CDC, and PcMab-47 was determined to be IgG 1 . In this study, we established the 47-mG 2a , a chimeric anti-PODXL antibody by combining variable region of PcMab-47 and constant region of mouse IgG 2a . We further produced 47-mG 2a -f, a core fucosedeficient 47-mG 2a to analyze antitumor activity in xenograft models [35].

Production of the mouse IgG 2a -type antibody PcMab-47
In this study, we first produced a mouse IgG 2a -type version of PcMab-47 by subcloning appropriate V H and V L cDNAs of PcMab-47 and C H and C L of mouse IgG 2a into pCAG vectors because mouse IgG 2a possess high ADCC and CDC activities ( Figure 1) [36]. The IgG 2a -type PcMab-47 was designated 47-mG 2a .

The binding affinity of mouse IgG 2a -type PcMab-47
We performed a kinetic analysis of the interactions of PcMab-47, chPcMab-47, 47-mG 2a , and 47-mG 2a -f with OSCC cells using flow cytometry. As shown in Figure
We stained 201 OSCC samples using PcMab-47 and 47-mG 2a and summarized our findings in Table 2 We also performed immunohistochemical analysis using PcMab-47 and 47-mG 2a against normal tongue, and checked the PODXL expression in normal squamous epithelium. As shown in Figure 6, PODXL was not detected in normal squamous epithelium. In contrast, PODXL was detected in normal endothelial cells. The staining intensity of 47-mG 2a was also higher than that of PcMab-47 in normal endothelial cells. We further checked the cross-reactivity of PcMab-47 against mouse PODXL using flow cytometry and immunohistochemistry. However, no reaction was observed against mouse PODXL (data not shown), indicating that anti-PODXL antibodies might not affect the tumor angiogenesis or tumor microenvironment in mouse xenograft model.

Functional analysis of PODXL in OSCC cells in vitro and in vivo
Next, we investigated whether PODXL is associated with tumor phenotype in OSCC cell lines. We selected SAS cells for this study because they express PODXL at higher levels than other OSCC cells ( Figure  3) and they have been reported to grow extremely well in 100 μL of serially diluted antibodies (6 ng/mL to 100 μg/mL), and secondary antibodies were then added. Fluorescence data were collected using a cell analyzer. GeoMean, geometric mean of fluorescence intensity. in vivo [43]. As shown in Figure 7A, PcMab-47 did not react with PODXL-knockout (KO) SAS cells (SAS/ hPODXL-KO). To examine the migratory and invasive abilities of SAS/hPODXL-KO cells, we performed wound-healing and invasion assays, respectively, but no significant differences in migration ( Figure 7B) and invasion ( Figure 7C) were identified between parental and SAS/hPODXL-KO cells. We next investigated whether PODXL is associated with the growth of OSCC cell lines in vitro using the MTS assay. The growth of three SAS/hPODXL-KO cell lines was lower than that of parental SAS cells ( Figure 7D). We further investigated whether PODXL affects OSCC tumor growth in vivo by comparing the growth of SAS and three SAS/hPODXL-KO cell lines that were transplanted subcutaneously into nude mice. As shown in Figure 7E, the growth of SAS/    hPODXL-KO cells was lower than that of parental SAS cells. Furthermore, we investigated the effect of PODXL on the growth of 3D cells with cancer stem cell-like properties. The 3D growth of three SAS/hPODXL-KO cell lines tended to decrease, and the growth of SAS/hPODXL-KO#23 cells was significantly inhibited after incubation for 72 h, indicating that PODXL has tumorigenic potential and might be associated with tumorigenicity in vivo ( Figure 8A and 8B). In contrast, there are little effects by anti-PODXL antibodies on 3D cell proliferation in vitro ( Figure 8C).

ADCC and CDC activities of 47-mG 2a and 47-mG 2a -f
We We also investigated the antitumor activity of 47-mG 2a and 47-mG 2a -f using HSC-2 xenograft models. 47-mG 2a and 47-mG 2a -f, and mouse IgG (control) were injected three times (100 μg of the antibodies on days 1, 7, and 14 after cell injections) into the peritoneal cavity of mice. Tumor formation was observed in all groups. 47-mG 2a -f significantly reduced tumor development compared with IgG on day 20 ( Figure 11A). Although 47-mG 2a also reduced tumor development compared with IgG, the result was not significant, indicating that the depletion of core fucose is critical for the antitumor activity of 47-mG 2a -f against HSC-2 xenograft models. The tumor weight of mice in the 47-mG 2a -f group was significantly lower than that in the IgG group ( Figure  11B). Conversely, 47-mG 2a was not associated with a significant decrease in tumor weight versus IgG. However, body weight was not significantly different among the three groups (Supplementary Figure 3A and 3B). The resected tumors of HSC-2 xenografts are depicted in Supplementary Figure 3C. No difference was illustrated among those groups using HE staining (Supplementary Figure 4).
We further investigated the antitumor activity of 47-mG 2a and 47-mG 2a -f using SAS xenograft models. In this study, we injected 100 μg of the antibodies three times. As shown in Figure 12A, 47-mG 2a -f, but not 47-mG 2a , significantly reduced tumor development compared with IgG on day 20, indicating that the depletion of core fucose is also critical for antitumor activity of 47-mG 2a -f against SAS xenograft models. Tumor weight was significantly lower in the 47-mG 2a -f group than in the IgG group ( Figure  12B). Conversely, 47-mG 2a was not linked to decreases in tumor weight. Body weight was similar among the three groups (Supplementary Figure 5A and 5B). The resected tumors of SAS xenografts are shown in Supplementary Figure 5C. No difference in HE staining was identified among the groups (Supplementary Figure 6).
We performed dose-escalation studies using CHO/ hPODXL xenograft models (Supplementary Figure  7). In this study, we injected 500 μg of the antibodies twice. As shown in Supplementary Figure 7A Figure 7C). Body weight was similar among the three groups (data not shown), as were HE staining patterns (Supplementary Figure 7D). In CHO/hPODXL xenograft models, doseescalation was not necessary because 100 μg of the antibodies twice showed enough anti-tumor activities.
We further performed dose-escalation studies using HSC-2 xenograft models because we could not observe www.oncotarget.com    . An asterisk indicates statistical significance ( * P < 0.05, Tukey-Kramer's test). www.oncotarget.com sufficient antitumor activity after three injections of 100 μg of the antibodies. In this study, we injected 500 μg of the antibodies twice. As shown in Figure 13A and 13B, both 47-mG 2a and 47-mG 2a -f reduced tumor development compared with control mouse IgG on day 15 (P < 0.01). Furthermore, 47-mG 2a -f exhibited greater antitumor activity than 47-mG 2a , but it is not significant ( Figure  13B). The resected tumors of HSC-2 xenografts are shown in Figure 14A. Tumor weights were significantly lower in the both 47-mG 2a and 47-mG 2a -f group than in the IgG group ( Figure 14B). Body weight was similar among the three groups (data not shown), as were HE staining patterns (Supplementary Figure 8). These results indicate that 47-mG 2a -f showed higher antitumor activities against HSC-2 due to its higher ADCC against PODXL-expressing OSCCs, and dose-escalation was effective for those anti-tumor activities in HSC-2 xenograft models.
We finally performed dose-escalation studies using SAS xenograft models because we could not observe sufficient antitumor activity after three injections of 100 μg of the antibodies (Supplementary Figure 9). In this study, we injected 500 μg of the antibodies three times. As shown in Supplementary Figure 9A Tumor weight was significantly lower in the 47-mG 2a -f group than in the IgG group, whereas 47-mG 2a did not reduce tumor weight (Supplementary Figure 9C). Body weight was similar among the three groups (data not shown), as were HE staining patterns (Supplementary Figure 9D). These results indicate that 47-mG 2a -f showed higher anti-tumor activities due to its higher ADCC against PODXL-expressing OSCCs, but dose-escalation was not sufficient for those anti-tumor activities in vivo.

DISCUSSION
47-mG 2a and 47-mG 2a -f showed higher binding activity compared with its original PcMab-47 ( Figure 4). We have sometimes experienced that chimeric antibodies possesses much higher affinity or lower affinity compared with original monoclonal antibodies [34,[44][45][46][47][48][49]. The stability of antibodies might be different among constant regions. Although 47-mG 2a exhibited ADCC activity against SAS cells ( Figure 9C), this activity is not sufficient to induce antitumor activity in vivo ( Figure 12). Then, we produced a non-fucosylated version of 47-mG 2a (47-mG 2a -f) to augment its ADCC activities because nonfucosylated antibodies are known to show higher ADCC activities [35,50]. As expected, 47-mG 2a -f exhibited stronger ADCC activities than 47-mG 2a against OSCC cells ( Figure 9B and 9C), leading to higher antitumor activities in HSC-2 and SAS xenograft models (Figure 11 and 12). In contrast, three injections were not effective despite a dose of 5 mg/kg, which corresponded to 100 μg/mouse in this study and exceeded the usual dose of antibody therapy [51,52]. Then, two injections of 500 μg/mouse, corresponding to 25 mg/kg, resulted in greater antitumor activities in HSC-2 xenograft models ( Figure  13), indicating that dose escalation might be necessary for monotherapy using anti-PODXL antibodies although  dose-escalation was not sufficient for SAS xenograft models (Supplementary Figure 9). Furthermore, we need to combine anti-PODXL antibodies with anti-cancer drugs or include them in novel antitumor regimens, including T cells and viruses, to exert antitumor activity against cancer cells. Previously, we produced chPcMab-47, a mouse-human chimeric antibody, which could be applied to cancer patients [34]. chPcMab-47 exerted ADCC and CDC activity and showed antitumor activities in mouse xenograft models. Because 47-mG 2a -f showed much higher ADCC activity in this study, we also need to produce core-fucose-deficient chPcMab-47 in the future study.
Jing et al. reported that high PODXL expression was significantly associated with worse OS and was predictive of shorter OS in multiple cancers, especially pancreatic and colorectal cancers [53]. They also revealed that high PODXL expression predicted worse DSS and DFS, although European patients were included in this analysis. These results suggest PODXL could be a prognostic factor, and diagnostic tools targeting this protein are expected. Because the association between PODXL expression and clinical stage has not been investigated in OSCC, it was analyzed in the current study. However, PODXL expression in stages III-IV was not significantly higher than that in stages I-II in patients with OSCC (data not shown). Further investigation regarding the clinical importance of PODXL expression in more patients with OSCC is needed.
Taken together, anti-PODXL antibodies could be useful antibody therapies against PODXL-expressing OSCCs. PODXL is known to be expressed in kidney, heart, pancreas, and breast tissues, and it plays important roles in those tissues [22]. Recently, we successfully produced cancer-specific mAbs (CasMabs) against human podoplanin [16,54]. Using the same methods, we need to develop CasMabs against PODXL in the near future.

Tissues
Cancer tissue microarrays of oral cancers were purchased from US Biomax (Rockville, MD, USA) and Cybrdi (Frederick, MD, USA). This study examined 53 patients with oral cancer who underwent surgery at Tokyo Medical and Dental University. The Tokyo Medical and Dental University Institutional Review Board reviewed and approved the use of human cancer tissues. Written informed consent was obtained for the use of human cancer tissue samples.

Antibodies
PcMab-47, a mouse anti-PODXL mAb (IgG 1 , kappa), was developed as previously described [17]. Mouse IgG was purchased from Sigma-Aldrich. To generate 47-mG 2a , appropriate V H and V L cDNAs of mouse PcMab-47 and C H and C L of mouse IgG 2a were subcloned into pCAG-Ble and pCAG-Neo vectors (Wako Pure Chemical Industries, Osaka, Japan), respectively. Antibody expression vectors were transfected into ExpiCHO-S cells using the ExpiCHO Expression System (Thermo Fisher Scientific). To generate 47-mG 2a -f, antibody expression vectors were also transfected into PDIS-5 cells using the ExpiCHO Expression System. Stable transfectants of chPcMab-47 cells were established previously [34].

Immunohistochemical analyses
Histologic sections (4 μm thick) were deparaffinized in xylene and then rehydrated and autoclaved in citrate buffer (pH 6.0; Agilent Technologies Inc.) for 20 min. Sections were then incubated with 0.5-5 μg/mL primary mAbs for 1 h at room temperature and then treated using an Envision+ kit (Agilent Technologies) for 30 min. Color was developed using 3,3-diaminobenzidine tetrahydrochloride (Agilent Technologies) for 2 min, and sections were then counterstained with hematoxylin (Wako Pure Chemical Industries Ltd.). The intensity of staining was evaluated as −, 1+, 2+, or 3+.

Proliferation assay in vitro
Cell proliferation in vitro was measured using MTS tetrazolium (Cell Titer 96 Aqueous, Promega, Madison, WI, USA). Cells were plated (500 cells/100 μL/well) in triplicate in 96-well plates. Cell viability was measured every 24 h for 96 h. After adding 20 μL of MTS to the wells followed by a 1-h incubation at 37°C, the absorbance at 490 nm (reference, 630 nm) was read using a microplate reader (Power Scan HT, Bio Tek Instruments, Winooski, VT, USA). The mean absorbance of the 3-well set was obtained from 0 to 96 h. Statistical significance was analyzed using the standard Student's t-test. P-values <0.05 were considered statistically significant.

Proliferation assay in vivo
Five-week-old female BALB/c nude mice were purchased from Charles River (Kanagawa, Japan). Sevenweek-old mice were used for the in vivo proliferation assay. Cells (0.3 mL of 1.33 × 10 8 /mL in DMEM) were mixed with 0.5 mL of BD Matrigel Matrix Growth Factor Reduced (BD Biosciences, San Jose, CA, USA). A 100 μL suspension (containing 5 × 10 6 cells) was injected subcutaneously into the left flanks of nude mice. The tumor diameter was measured using calipers, and the tumor volume was calculated using the following formula: volume = W 2 × L/2, where W is the short diameter and L is the long diameter. The mice were euthanized 21 days after cell implantation. All data were expressed as the mean ± SEM. Statistical analysis was performed using a twotailed Student's t-test. P-values <0.05 were considered statistically significant.

Invasion assay
The invasion assay was performed using a Cytoselect 96-well Collagen Cell Invasion Assay Kit (Cell Biolabs, San Diego, CA, USA) according to the manufacturer's instructions. Cells (2.5 × 10 4 ) were preincubated in DMEM containing 0.5% FBS for 12 h, washed with serum-free DMEM, and added to the top insert of a chamber containing a polycarbonate membrane with 8-μm pores coated with a layer of bovine type I collagen matrix. The lower chamber was filled with DMEM containing 10% FBS as the chemoattractant. The assembled chambers were then placed in the CO 2 incubator for 20 h. Cells that passed through the membrane pores were lysed and quantitated using a www.oncotarget.com fluorescence dye-containing solution and a fluorescence plate reader. Statistical significance was analyzed using the standard Student's t-test. P-values <0.05 were considered statistically significant.

Wound-healing assay
Cells were grown to confluence in DMEM containing 0.5% FBS in 6-well plates. Scratch wounds were made using 10-μL sterile pipette tips. The cells were incubated in DMEM containing 10% FBS for the indicated times. Images were captured using an Evolution MP camera (Media Cybernetics, Rockville, MD, USA).

3D cell proliferation assay
3D cell proliferation in vitro was measured using the CellTiter-Glo ® 3D cell viability assay (Promega) according to the manufacturer's instructions [55]. Briefly, the cells were plated (500 cells/100 μL/well) in triplicate in 96-well ultra low attachment plates (Costar Corning, Schiphol-Rijk, Netherlands) with DMEM containing 10% FBS. The 3D cell viability was measured every 24 h for 72 h. The CellTiter-Glo ® 3D reagent was added into wells in a 1:1 dilution (100 μL volume in well: 100 μL of reagent) and then the solutions were mixed well by pipetting. After incubation for 30 min at 37°C, the luminescent signal was read using an EnSpire multi-plate reader (Perkin Elmer, PerkinElmer, Waltham, MA, USA). Images were captured using an Evolution MP camera (Media Cybernetics). In addition, to investigate the influence of the antibodies on the 3D cell proliferation, the cells were plated as described above with or without the antibodies (30 μg/ ml). The proliferation rate was calculated relative to the signal at 0 h.

ADCC
ADCC was examined using the 51 Cr-release assay. Effector cells were prepared as previously described [36]. Mouse splenocytes were harvested from spleens of sixweek-old male SCID mice. Spleens were homogenized in RPMI 1640 and centrifuged. To deplete red blood cells, the cell pellet was suspended in red blood cell lysis buffer (Sigma-Aldrich). After washing and re-suspension in 10% FBS/RPMI1640, splenocytes were used as effector cells. Target cells were incubated with 0.1 μCi of [ 51 Cr]sodium chromate at 37°C for 1 h. After three washes with 10% FBS/RPMI1640, 51 Cr-labeled target cells were placed in 96-well plates in triplicate. Effector cells and antibodies (10 μg/ml) were added to the plates (E/T ratio = 100). After 6, 12, and 18 hrs of incubation, 51 Cr release was measured in the supernatant (100 μL) from each well using a gamma counter (PerkinElmer). The percentage of cytotoxicity was calculated using the following formula: % specific lysis = (E -S)/(M − S) × 100, where E is the 51 Cr release in the test sample, S is the spontaneous release, and M is the maximum release. Statistical significance was analyzed using the standard Student's t-test. P-values <0.05 were considered statistically significant.

CDC
CDC was evaluated using the 51 Cr release assay. Target cells were incubated with [ 51 Cr]sodium chromate (0.1 μCi) for 1 h at 37°C. The cells were then washed with 10% FBS/RPMI1640. The 51 Cr-labeled cells were incubated with a baby rabbit complement Cedarlane (Ontario, Canada) at a dilution of 1:4 in the presence of antibodies (10 μg/mL) for 6 h in 96-well plates. After the incubation, 51 Cr radioactivity was measured in the supernatants using a gamma counter. The percent cytotoxicity was calculated using the following formula: % specific lysis = (E -S)/(M − S) × 100, where E is the release in the test sample, S is the spontaneous release, and M is the maximum release. Statistical significance was analyzed using the standard Student's t-test. P-values < 0.05 were considered statistically significant.

Antitumor activity of anti-PODXL antibodies
Five-week-old female BALB/c nude mice were purchased from Charles River and used in experiments at 7 weeks of age. Cells (0.3 mL of 1.33 × 10 8 /mL in DMEM) were mixed with 0.5 mL of BD Matrigel Matrix Growth Factor Reduced (BD Biosciences). A 100-μL suspension (containing 5 × 10 6 cells) was injected subcutaneously into the right flanks of nude mice. After 1 day, 100 μg or 500 μg of 47-mG 2a , 47-mG 2a -f, or mouse IgG in 200 μL PBS were injected into the peritoneal cavity of each mouse. Additional antibodies were injected once weekly for several weeks. The tumor diameter and tumor volume were determined as previously described. The mice were euthanized 15, 16, 20, or 21 days after cell implantation. All data were expressed as the mean ± SEM. Statistical analysis was performed using the Tukey-Kramer test. P-values < 0.05 was considered statistically significant.