According to the previous report for OAcGD2 detection by IHC (16), frozen section of the tumor must be used for IHC of OAcGD2 as the deparaffination process may leach out hydrophobic glycolipid molecules such as OAcGD2. Since our original breast cancer specimens are available only as paraffin-embedded tissues, it is difficult to assess OAcGD2 expression in primary tumors. Fortunately, in recent years, the focus of the CSC field has shifted to the use of freshly isolated tumor specimens and early-passage patient-derived xenografts (PDXs), instead of using cultured tumor cell lines (21). Xenotransplantation assays have become an important strategy to assess CSC subpopulations and their activities. We have established five breast cancer PDXs with various molecular subtypes including luminal A and B, triple-negative breast cancer ( Supplemental Table 1 ), and identified ALDH as a BCSC marker for BC0244, BC0634, BC0350, VBC108, and CD44⁺CD24⁻ as BCSC marker for BC0145 PDXs ( Table 1 ) (22–25). In addition, ALDH is identified as a BCSC marker for PDX AS-B244, which was a subclone of BC0244, designated as AS-B244 (25). Flow cytometry analysis showed that 13–30% of the PDX cells expressed the indicated BCSC markers ( Table 1 ). Examination of their expression of OAcGD2 by flow cytometry with anti-OAcGD2, mAb8B6, showed that 30–100% of BCSCs in these six PDXs expressed OAcGD2. Furthermore, Pearson correlation analysis of BCSCs and OAcGD2 expression in these PDXs showed positive correlation of BCSC percentage with OAcGD2 MFI (r=0.8115, p=0.05), and percentage of OAcGD2⁺ in BCSCs (r=0.85, p=0.03), but not with percentage of OAcGD2⁺ cells (r=0.42, p=0.41) ( Figure 1A ). Thus, the amount (MFI) of OAcGD2 is much more pertinent to stemness than its percentage. These findings suggest that OAcGD2 may serve as a marker for further enrichment of BCSCs.

The hallmark of CSCs is their ability to initiate tumors better than their bulk tumor counterparts (26). To determine whether OAcGD2-expressing ALDH⁺ or CD44⁺CD24⁻ BCSCs are more tumorigenic than OAcGD2 negative ALDH⁺ or CD44⁺CD24⁻ BCSCs, we sorted the highest and lowest OAcGD2-expressing ALDH⁺ BC0244 ( Figure 1B ) or CD44⁺CD24⁻ BC0145 cells ( Figure 1C ) to assess their mammosphere-forming ability. The OAcGD2⁺ALDH⁺ BC0244 and OAcGD2⁺CD44⁺CD24⁻ BC0145 cells formed more mammospheres when compared with OAcGD2⁻ALDH⁺ BC0244 (40.2 ± 4.4 vs. 21.5 ± 2.1) and OAcGD2^-CD44⁺CD24⁻ BC0145 cells (30.0 ± 3.3 vs. 15.1 ± 1.7) (p < 0.05 for both) ( Figures 1D, E ). To determine the tumor-initiating potentials of OAcGD2⁺ BCSCs, we sorted OAcGD2^low and OAcGD2^high BCSCs from BC0244 and BC0145 cells and injected these cells into the mammary fat pad of NOD/SCID mice (n = 3/group) at the cell doses of 10², 10³, and 10⁴. As shown in Table 2 , OAcGD2^low subpopulation of BC0244 and BC0145 failed to show any tumor engraftment at all three cell dose levels. In contrast, two out of three mice inoculated with 10³ or 10⁴ OAcGD2^high BC0244 BCSCs grew tumor. Similar results were found in mice injected with OAcGD2^high BC0145 BCSCs. These findings support the notion that OAcGD2 is a marker for BCSCs.

Next, we evaluated the antitumor efficacy of mAb8B6, an anti-OAcGD2, on two PDXs of breast cancer, BC0244 and BC0145. Tumor cells were injected into the mammary fat pad of NOD/SCID mice. When the tumor volume reached 100 mm³, mice were treated with mAb8B6 (5 mg/Kg), 2 mg/kg of doxorubicin (DOX-high), 0.5 mg/kg of doxorubicin (DOX-low), or PBS once a week for 4 weeks. All animals were sacrificed 3 days after the last administration. As shown in Figure 2A , the growth of BC0244 and BC0145 tumors of individual mice and the slope of tumor growth for each group were significantly suppressed by DOX and mAb8B6, with mAb8B6 compared to PBS, p<0.0001 in BC0244 and p=0.0012 in BC0145. Treatment with DOX showed dose-dependent inhibition-(DOX-high vs. DOX-low, p=0.0039 in BC0244, p=0.0236 in BC0145), and mAb8B6 was as effective as DOX-high in tumor growth inhibition (p=0.4195). At the time of sacrifice, the tumor volume of BC0244 in the DOX-low, DOX-high, and mAb8B6 groups was significantly reduced to 56.0 ± 8.6%, 34.1 ± 4.9%, and 38.7 ± 6.5%, respectively, of the control group treated with PBS (p<0.001), although the reduction in tumor volume of BC0145 was significant only in the mAb8B6 group (25.8 ± 7.4% of the PBS, p=0.0012), but not in the DOX-low (60.5 ± 12.5%), or DOX-high 43.3 ± 23.9% ( Figure 2B ). Thus, based on tumor growth rate or tumor volume harvested after treatment, anti-OAcGD2 treatment significantly suppressed the tumor growth of both PDXs. In addition, we examined the BCSCs (ALDH⁺ for BC0244 and CD44⁺ CD24⁻ for BC0145) in the harvested tumors. BCSCs in both mAb8B6 treated BC0244 and BC0145 tumors decreased to 80.0 ± 17.9% (p=0.049) and 68.5 ± 15.6% (p=0.0017), respectively, of PBS control group ( Figure 2C ). Although DOX at 2 mg/kg reduced tumor volume, it had no effect on BCSCs when compared to the PBS control group. Interestingly, treatment with 0.5 mg/kg of DOX increased the BCSCs in both BC0145 and BC0244 tumors to 130.9 ± 19.7% (p=0.006) and 127.2 ± 26.7% (p=0.034) of control group, consistent with the reported relative resistance of BCSCs to chemotherapy.

To further confirm that anti-OAcGD2 treatment can target BCSCs, we tested the frequency of tumor-initiating cells by limiting dilution engraftment assay. BC0244 and BC0145 tumor cells isolated from the mAb8B6- and DOX-treated mice were inoculated into NOD.SCID mice at three different cell doses: 10⁴, 10⁵, and 10⁶. Tumor formation was monitored for 2 months ( Table 3 ). As expected, the frequency of tumor-initiating cells in BC0244 tumor treated with mAb8B6 (1:94,752) was significantly lower than those treated with PBS (1:14,241), DOX-low (1:9,100), and DOX-high (1:14,241). Similarly, in BC0145-bearing mice, the frequency of tumor-initiating cells from the mAb8B6 group (1:66,954) was much lower than those treated with PBS (1:14,241), DOX-low (1:5,581), and DOX-high (1:21,636) groups. In line with the increased percent of BCSCs as determined by surface markers, treatment of the BC0244 and BC0145 tumor-bearing mice with 0.5 mg/kg DOX increased the frequency of tumor-initiating cells.

We further investigate the in vitro effect of the anti-OAcGD2 on the proliferation of OAcGD2^high and OAcGD2^low BC0145/BC0244, as determined by AlamarBlue assay. As shown in Figure 3A , there was no difference in the proliferation rate between OAcGD2^high and OAcGD2^low subpopulations of both PDXs, but treatment with mAb8B6 inhibited proliferation of OAcGD2^high BC0145/BC0244 cells only (left panel), not OAcGD2^low BC0145/BC0244 cells (right panel). We next examined the effects of mAb8B6 on the properties of BCSCs. Mammosphere-forming capacity of sorted ALDH⁺OAcGD2^low and ALDH⁺OAcGD2^high BC0244 cells was assessed in the absence/presence of mAb8B6 at 25 or 50μg/ml. The presence of mAb8B6 significantly decreased mammosphere formation of ALDH⁺OAcGD2⁺ in a dose-dependent manner, while only slightly attenuated the mammosphere formation of ALDH⁺OAcGD2^low cells ( Figures 3B, C ).

It has been reported that mAb8B6 inhibited the growth of neuroblastoma, small cell lung cancer, and lymphoma cell lines, which was mediated by ADCC/CDC and induction of apoptosis (16, 27). To determine whether mAb8B6 exerts direct cytotoxicity on breast cancer, we examined apoptosis of BC0244 after incubation with mAb8b6 or isotype control antibody for 24 h by flow cytometry. As shown in Figure 4A , mAb8B6 induced greater early apoptosis (16.7 and 21.8% at 10 and 50 μg/ml, respectively) and late apoptosis (8.3 and 11.1% at 10 and 50 μg/ml, respectively) of OAcGD2^high BC0244 cells as compared to the isotype control antibody (early apoptosis: 4.9%; late apoptosis: 0.7% at 50 μg/ml). On the other hand, mAb8B6 did not induce obvious apoptosis of OAcGD2^low BC0244 cells. These results demonstrated the ability of mAb8B6 in inducing programed cell death in OAcGD2-expressing cells.

To further ascertain whether mAb8B6 can inhibit tumor growth of OAcGD2-expressing cells in vivo, we inoculated 1x10⁵ OAcGD2^high BC0244 cells into NOD/SCID mice. When the tumor volume reached 100 mm³, mice were randomly divided into two groups for treatment with mAb8B6, or PBS control every week by i.v. injection. It is noteworthy that 50% of mAb8B6 (3 of 6) were completely tumor-free. At 4 weeks, tumors of the remaining three mice of the mAb8B6 group were reduced to 9.4 ± 2.5% (p<0.001) of PBS control ( Figure 4B ). Moreover, the remaining tumors from mice treated with mAb8B6 contained significantly less ALDH⁺ BCSCs when compared to those treated with PBS (33.3 ± 15.7% of PBS control, p=0.009) ( Figure 4C ). To determine whether the anti-proliferative and apoptotic activities of mAb8B6 observed in vitro is mimicked in vivo, we examined the percentage of Ki67⁺ cells in harvested tumors ( Figure 4D ). The Ki67⁺ cell of tumors obtained from mice treated with mAb8B6 was 12.3± 4.0%, which was significantly lower than the PBS control group (49.8 ± 8.2%, p = 0.002). TUNEL-staining revealed extremely low levels of apoptosis in the tumors from PBS-treated mice (5.0 ± 2.0%) ( Figure 4E ). The percentage of apoptotic cells was significantly higher in the tumors from mice treated with mAb8B6 (66.3 ± 11.6%; p = 0.002). These findings indicate that mAb8B6 can target CSCs by inducing apoptosis and suppress tumor growth as illustrated in Figure 5 .

Human clinical breast cancer specimens were obtained from patients at the time of initial surgery and were fully encoded to protect patient confidentiality. Clinical specimens were utilized under a protocol approved by the Institutional Review Board of the Human Subjects Research Ethics Committee of Academia Sinica, Tri-Service General Hospital, and Veterans General Hospital (Taipei, Taiwan). Isolation of the primary tumor cells from clinical specimens was described previously (24). Five patient-derived xenografts (PDXs) were successfully established from patients BC0145, BC0244, BC0350, and BC0634. BCSC subpopulation was delineated as CD24⁻CD44⁺ cells in BC0145, and ALDH⁺ cells in BC0244, BC350, and BC0634, according to their tumorigenicity (22, 25). All PDXs were maintained throughout xenograft passages. Monolayer cultures of H-2Kd⁻ALDH⁺ BC0244, sorted from xenograft tumors of human primary breast cancer, were designated as AS-B244 cells as described previously (22, 25). The anti-OAcGD2 mAb8B6 is kindly provided by OGD2 Pharma, France.

Cell surface OAcGD2 expression on tumor cell lines was assessed by indirect immunofluorescence. Cells were incubated with either mAb8B6 or mouse IgG3 (isotype control antibody) at 10 μg/ml for 30 min at 4°C in 0.1% BSA-PBS. After the reaction, these cells were incubated with the FITC-conjugated goat anti-mouse IgG as a second antibody (Biolegend) for 30 min at 4°C. BCSCs were defined as CD44⁺CD24⁻ or ALDH⁺ cells. ALDH activity was determined by the ALDEFLUOR™ kit (Stem Cell Technologies) according to the manufacturer’s instructions. All stained cells were then examined by EC800 flow cytometer (SONY). For sorting, the cells were collected using a BD FACSAria II flow cytometer (BD Biosciences).

The sphere culture was performed as previously described (47) with some modifications. OAcGD2⁻ or OAcGD2⁺ CSCs from BC0244 or BC0145 cells were FACS sorted using antibodies against OAcGD2, ALDH, CD44, and CD24. The sorted cells (1x10³) were incubated in a mammosphere growth medium in ultra-low-attachment 96-well plates (Corning). All cells grew at a density of 1×10⁴ cells/ml in serum-free Dulbecco’s Modified Eagle’s Medium/F12 supplemented with 20 ng/ml epidermal growth factor, 10 ng/ml basic fibroblast growth factor, 5 µg/ml insulin, 0.4% bovine serum albumin, 100 U/ml Pen/Strep, and 2% B27. Monoclonal antibodies were diluted and added to each well containing 1,000 tumor cells in 96-well plates to give the final concentrations of 0, 25, and 50 μg/ml. After 7 days, the resulting mammospheres were counted.

Cell viability was measured using the Alamarblue assay. Briefly, sorted cells were incubated with/without mAb8B6 (50 μg/ml) for 72 h at 37°C with 5% CO₂. Absorbance was measure at 570 nm on a SpectraMAX (M3). The proliferation rate was calculated by normalizing to 0 h.

Cells (2×105 cells) were plated in six-well plates for 24 h at 37°C in a humidified atmosphere containing 5% CO₂, and then treated with 50 μg/ml of mAb8B6 and mIgG for 24 h. After incubation, cells were stained with FITC-conjugated goat anti-mouse IgG (Jackson) as described above. After washing twice with PBS, we resuspended these cells in 500 µl of a binding buffer with Annexin V in the dark for 20 min, according to the manufacturer’s protocol (BD).

NOD/SCID mice were purchased from Jackson Lab and maintained at the animal facility of the Chang Gung University (IACUC number: CGU106-055). Animal studies were conducted by the guidelines for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee. BC0244 and BC0145 cells (5×10⁵) mixed with 100 µl of 2 mg/ml Matrigel were injected at the base of the nipple of the fourth abdominal fat pad of female mice (4- to 6-week-old). Ear numbering system was used to create a unique identifier; the tumor-free mouse was exclusive. When tumor size reached 100 mm³, mice (n = 6/group) were randomly assigned to each group and i.v. injected with 5 mg/kg of mAb8B6, 2 mg/kg of doxorubicin (DOX-high), 0.5 mg/kg of doxorubicin (DOX-low), and PBS once a week for 4 weeks. Sorted OAcGD2⁺ALDH⁺ BC0244 cells (1×10⁵) mixed with 100 µl of 2 mg/ml Matrigel were injected at the base of the nipple of the fourth abdominal fat pad of female mice. When tumor size reached 100 mm³, mice (n = 6/group) were randomly assigned to each group and treated with 5 mg/kg of mAb8B6 or PBS once a week for 4 weeks with or without PBMC intraperitoneally at 1×10⁷/mouse as effector cell. Tumor volume was monitored using a vernier caliper twice a week for up to 8 weeks and calculated according to the equation: V = 1/2*W²*L, where L is the length and W the width of a tumor. For ethical considerations, mice had to be euthanized once tumor volume had reached 2,000 mm³, which was considered the endpoint for each mouse. In addition, OAcGD2^high of ALDH⁺ cells from BC0244 cells and OAcGD2^high of CD44⁺CD24^- cells from BC0145 cells were sorted using FACSAria II cell sorter (BD) and then inoculated in NOD/SCID mice (n=5). These mice were treated with mAb8B6 or PBS as described above.

To obtain single cells from the tumors, we sliced a tumor into square fragments of 1 mm² and then digested these fragments by incubation in a MEM medium containing collagenase (1,000 U/ml), hyaluronidase (300 U/ml), and DNase I (100 μg/ml) at 37°C for 1 h. Single cells (10⁴, 10⁵, or 10⁶) isolated from tumor-bearing mice treated with mAb8B6, DOX-high, DOX-low, or PBS were injected at the mammary fat pad. Animals were examined for tumor formation after 1 week. The frequency of tumorigenic cells and the 95% confidence interval were calculated using Extreme Limiting Dilution Analysis (48).

Tumor tissue sections were deparaffinized followed by antigen retrieval by autoclave for 121°C, 5 min in AR-10 solution (Biogenex). Endogenous peroxidase was quenched before saturating with H₂O₂ blocking solution (Dako). Sections were stained with mouse anti-human Ki67 mAb (Leica, Cat No. NCL-L-Ki67-MM1). Bound antibody was detected by polymer-HRP IHC detection system (Biogenex). Digital images were captured by Aperio ScanScope XT Slide Scanner (Aperio Technologies, Vista, CA, USA) under 20× magnification. Positive and negative stained cells were counted on five random fields for each tumor. Data were expressed as cells positive for Ki67 staining/total cells. 

The extent of apoptosis in the tumors was measured by TUNEL using the TUNEL assay kit (Abcam, ab206386) following the manufacturer’s protocol. Data were expressed as cells positive for TUNEL staining/total cells.

Statistical analysis was performed using Prism (GraphPad Software). All values are presented as means ± SD. Three independent experiments were performed, and representative results were shown. * p < 0.05, ** p < 0.01, *** p < 0.001. P-value was calculated by using the Student t-test or one-way ANOVA.