Reagents

OvCa cell lines OVCAR3, OV90, and SKOV3 cell lines were obtained from ATCC. OVCAR4 was generously provided by Dr. Panos Konstantinopoulos (Dana Farber Cancer Institute, Harvard Medical School, Boston, MA). OVCAR3 cells were cultured in RPMI 1640 medium (Cat #11,875,093, Gibco-BRL, Gaithersburg, MD) supplemented with 10% fetal bovine serum (FBS, Cat #26,140,079, Thermo Fisher Scientific, Waltham, MA), 1X penicillin-streptomycin (Cat # 15,070,063, Thermo Fisher Scientific), and 0.01 mg/ml bovine insulin (Cat #I0516, Sigma-Aldrich, Natick, MA). OVCAR4 cells were cultured in RPMI 1640 supplemented with 10% FBS and 1X penicillin-streptomycin (Life Technologies, Carlsbad, CA). OV90 cells were cultured in a 1:1 mixture of MCDB 105 medium (Cat #117,500, Cell Applications, San Diego, CA) and Medium 199 (Cat #11,150,059, Gibco-BRL) and 10% FBS. These cell lines have been characterized previously [20,21,22]. SKOV3 Control (SKOV3CTL and SKOV3 cells overexpressing ST6GALNAC1 enzyme which results in elevated STn (SKOV3-ST6GALNAC1), described in Prendergast et al. [6], were grown in McCoy’s media (Cat # 16,600,082, Gibco-BRL) supplemented with 10% FBS and 1X penicillin and streptomycin. All cell lines were maintained at 37˚C in 5% CO2. Established cell lines were subjected to human cell identity verification (STR profiling) at Dana Farber Cancer Institute (http://moleculardiagnosticscore.dana-farber.org). All established cell lines were regularly tested for mycoplasma contamination (Lonza MycoAlert® Mycoplasma Detection Kit Cat # LT07-418, Walkersville, MD).

To better understand the unknown levels of STn in serum, it was necessary to confidently determine the levels of STn on a known control protein, to this end we utilized bovine submaxillary mucin (BSM). BSM is a glycoprotein that is rich in STn and can be used to extrapolate unknown sera STn expression. The lot of BSM used SLBH5656V (Sigma-Aldrich, Cat # M3895) was determined to have 14% sialic acids bound (provided by vendor). This is 0.14 µg sialic acid per µg of BSM. Based on the known sialic modifications of BSM we assumed this is 0.14 µg STn per 1 µg of BSM.

Antibody generation

The generation of the murine and humanized anti-STn and anti-STn-ADC antibodies used in this assay have been described in the literature by Eavarone et al. [4] and Prendergast et al. [6].

MTT cytotoxicity assay

OVCAR3, OVCAR4, OV90, SKOV3CTL, or SKOV3-ST6GALNAC1 cells were seeded in 96-well plates, incubated overnight in the appropriate complete culture medium, then treated with increasing doses of the human anti-STn-ADC at 0, 2.5, 5, 10, and 50 or 100 nM and then incubated for 72 h or 6–7 days. Cell viability was determined by MTT assay (Cat # M6494, Thermo Fisher Scientific), and the percentage was calculated relative to control (vehicle treated) samples using the formula = (OD sample / OD of control average) × 100.

Flow cytometry

Flow cytometry was used to assess STn levels in cell lines and primary tumor cells. Following trypsinization, or tissue processing, and incubation with FcR blocking reagent (Miltenyi Biotec, Bergisch Gladbach, Germany, Mouse, Cat # 13,009,257), cells were stained with highly specific mouse anti-STn antibody (Siamab Therapeutics, Inc., Newton, MA) directly conjugated to Alexa Flour 647 using the Zenon antibody labeling kit (Thermo Fisher Scientific Cat # Z25108) or labeled with Alexa Flour 488 (Invitrogen, Carlsbad, CA, Cat # A11017). Fixable Live/Dead Violet (Thermo Fisher Scientific Cat # L34955) was used to determine the level of live and dead cells to exclude dead cells from the staining analysis. After washing, cells were fixed in 4% paraformaldehyde for 20 min, washed and reconstituted in PBS, and then analyzed using Guava Millipore (Millipore, Burlington, MA) or LSRII). Data were analyzed using FlowJo software (version 10.0.8). For analysis of cells derived from xenograft tumors, PDX tumor cells were stained with H2kD (BD Biosciences, Franklin Lakes, NJ) to exclude mouse cells from the sample at the time of analysis and only determine levels of STn expression in human cells.

In vivo treatment studies

All mouse studies were carried out in compliance with our Institutional Animal Care and Use Committee guidelines at Massachusetts General Hospital. For establishment of PDXs, tumor cells originally derived from patients (who provided informed consent) diagnosed with HGSOC were processed to homogenized single cell suspension and re-suspended in PBS: Matrigel ® (Cat # 354,234, Sigma-Aldrich) (1:1) and subcutaneously (s.c.) injected into ∽ 8-week-old female NOD/SCID mice (Jackson Laboratory, Bar Harbor, ME). Once initial tumor xenografts were formed, they were harvested for histological verification and tumor was cryopreserved for future rederivation. For the present study cryopreserved tumor samples were thawed and implanted in mice and allowed to form. Once sufficient tumor burden was formed the tumors were harvested, mechanically and enzymatically processed to remove mouse cells and an equal number of tumor cells were resuspended and injected into immunocompromised mice as described above and previously [23,24,25].

All animals were monitored regularly for tumor formation, and tumor volume was calculated using the formula (length x width x height)/2 as has been described [26]. When tumor volumes averaged between 200 and 250 mm3 the mice were randomized into the different arms to have a similar average tumor volume in each arm (a minimum of 4 mice/arm). Mice were treated with either vehicle control, isotype-ADC, human anti-STn-ADC, carboplatin/paclitaxel (Cat # C2538/ Cat # T1912 Sigma Aldrich) or a combination thereof, depending on the experimental design. Human anti-STn-ADC was administered at 5 mg/kg in sterile saline weekly as determined by our previous pharmacokinetic studies [4]. Similarly, the animals treated with isotype-ADC were also treated at 5 mg/kg in sterile saline weekly. Carboplatin and paclitaxel treatments were at 25 mg/kg and 12 mg/kg, respectively, weekly. Carboplatin and paclitaxel dose concentrations were based on our prior studies [23, 25]. Tumors were measured every 3 to 4 days with calipers, and mice were weighed twice a week. If mice lost more than 15% of their body weight or tumors grew to exceed the preset limitations, they were sacrificed. At the study’s completion, tumor samples from some xenografts were formaldehyde-fixed and paraffin-embedded for STn staining.

Organoid culture and treatment

Organoids were cultured as described previously [27]. All organoids were tested for mycoplasma (MycoAlert® Kit) and all were negative. For sensitivity analysis, organoids were digested with TrypLE (Life Technologies Cat #12,604), plated in 20% Matrigel, and then treated with either human anti-STn-ADC or Isotype-ADC at 0, 0.02, 0.05, 0.1, 0.15, 0.2, or 0.25 nM and read by CellTiter-glo (Cat # G7572, Promega, Madison, WI) six days later. On day 1, four wells were read with CellTiter-glo for growth rate correction. Growth rate corrected dose curves and area over the curve were calculated as described [28].

Patient selection and data

Blood samples were collected from patients with either OvCa or benign masses and processed for serum. Patient consent forms and sample collections were approved and in compliance with the Institutional Review Board guidelines (Protocol #s 07–049, 2016P002742 and 2000P001678). Research coordinators double-coded the samples, and the samples were run blindly. At the end of the study, the code was revealed, and results analyzed as described in the statistical analysis section. Patient characteristics are shown in Supplemental Table 1. Four hundred twenty serum samples were obtained from the VCRB MGH Gyn sample repository. All samples that were used for diagnostic assessment were collected within five days prior to surgical debulking. Clinical CA-125 values were obtained from the patient’s chart. If there were multiple CA-125 values, the results closest to the time serum was collected before surgery were used for comparison. A representative tumor piece was fixed in paraformaldehyde for a subgroup of patients and used for immunohistochemistry studies described below. Clinical correlates such as histologic subtype, stage, grade, first line treatment, number of treatment cycles, and response to treatment were collected (see Supplemental Table I).

Independent of the diagnostic potential, we initiated a preliminary experiment to assess what impact surgical debulking had on circulating STn levels. For this we obtained a set of matched blood samples from patients diagnosed with HGSOC prior to surgery and again post-surgery and subjected them to our ELISA as described.

Immunohistochemistry (IHC)

A subset of paraffin-embedded ovarian tumor samples from patients with matched blood sampled were sectioned at 5 μm thickness to compare tissue levels of STn with the patients corresponding serum STn levels. The subset of matched samples included 41 serous, 10 mucinous, 12 clear cell, and 4 endometrioid histologies. Antigen retrieval was performed using 10nM sodium citrate solution at 120º C using a pressure cooker for 15 min. Tissues were incubated with 3% H2O2 (Cat # S25359, Thermo Fisher) for 20 min, then blocked with 6% serum cocktail (normal Horse Cat # S2000, Bovine Cat # SP5050, Goat Cat # S1000 serum from Vector Labs, Burlingame, CA) for 20 min. Tissues were then incubated with either murine anti-STn primary antibody or control mouse antibody MOPC isotype at 10 µg/mL (MOPC173, Biolegend, San Diego, CA), or no primary antibody diluted in the blocking cocktail. After washing with PBST, slides were incubated for 45 min with anti-mouse antibody (Santa Cruz, Dallas, TX). To visualize the staining, 3,3’-diaminobenzidine (DAB) was used (Cat # SK-4100; Vector stains, Burlingame, CA). Slides were counterstained with hematoxylin (Cat # CS402-1D Thermo Fisher Scientific) and Scott’s water. Finally, slides were dehydrated and mounted with coverslips. Membrane STn positivity was scored by a board-certified pathologist (SJH) blinded to the sample code. A score of 0 meant minimal membrane staining with STn in less than 5% tumor cells, or no staining with STn was identified. A score of 1 indicated minimal to moderate STn membrane staining in 5–24% of tumor cells. A score of 2 indicated moderate to strong membrane staining in 25 − 50% of tumor cells. A score of 3 indicated moderate to strong membrane staining in greater than 50% of the tumor cells.

Development of an STn biomarker ELISA screening assay

To develop a sandwich ELISA to identify STn expression in patient serum, we utilized commercial, and Siamab generated murine anti-STn antibodies [4, 6] in a research use only (RUO) grade assay. A subset of these anti-STn antibodies was chosen for sandwich ELISA testing, and pairing was guided based on their light chain CDR sequences. Proof of concept studies matched murine and humanized antibody pairs. The pairing included murine S3F, 2G12-2B2, CC49, 5G2-1B3, B72.3, and humanized Hu3F1 L1H1, Hu2G12-2B2 L0H3, Hu5G2-1B3 L1H2 for testing along with isotype controls in a matrix format. Isotype controls included were murine MOPC173 (BioLegend, Cat # 400,264), Plates (Corning, Cat # 9018) were coated with 1, 3, or 5 ug/mL murine antibodies in coating buffer (50 mM sodium carbonate/bicarbonate pH 9.5) overnight at 4 C. After 3x washing with phosphate-buffered saline with 0.05% Tween-20 (PBS-T), plates were blocked with blocking buffer (1% ovalbumin (OVA) in PBS) for 1 h at room temperature. Buffer was removed, and 100 µl/well of STn glycoprotein sample (0.000125 mg/mL BSM, Sigma Aldrich) diluted in blocking buffer was added and incubated for 90 min at 37ºC. Plates were then washed 2x with PBS and a subset of wells had their sialic acids oxidized by treatment with 2 mM periodate. Wells were incubated for 20 min at 4 °C. Plates were then washed 3x with PBS-T, and 3ug/mL secondary humanized anti-STn antibodies were added to the wells diluted in blocking buffer and incubated for 1 h at room temperature. Plates were washed 3x with PBS-T followed by incubation with 0.08 µg/mL peroxidase-conjugated goat anti-human antibody (Cat #109-035-098, Jackson ImmunoResearch, West Grove, PA) for 1 h. Next, wells were washed 3x with PBS-T, and then wells were incubated with 100 µL of enzyme substrate (0.5 mg/mL o-phenylenediamine; 0.03% H2O2 in citric/phosphate buffer pH 5.5). The enzyme reaction was terminated by the addition of an equal volume of 1.6 M sulfuric acid. Optical Density (OD) readings of periodate and non-periodate treated wells were determined at 490 nm. Binding affinities of the antibody pairs were compared by subtracting the periodate treated wells from the non-periodate treated wells to obtain the periodate-sensitive STn binding.

Statistical analysis

All experiments were carried out as at least 3 biological replicates and the data were analyzed with GraphPad Prism (GraphPad Software, La Jolla, CA). Patient serum biomarkers were analyzed using the R statistical computing environment (4.0.2) [29] and Stan, a platform for statistical modeling [30] using the R package rstan [31]. Bars represent mean ± SEM. One-way ANOVA was conducted to assess for significant differences seen in tumors treated with human anti-STn-ADC. A statistical significance is determined if the p value was < 0.05.

To assess the complementarity of STn to CA-125 as a biomarker for ovarian cancer, bivariate mixture models [32] estimated the joint distribution of STn and CA-125. The underlying distribution was a bivariate t-distribution to provide robustness to outliers. A mixture model for one serum biomarker (e.g., CA-125) in cases has a fraction of cases (∽ 80%) that overexpress the biomarker compared to control patients, and the complementary fraction (∽ 20%) that does not overexpress the biomarker, that is, its distribution is the same as the distribution for the control group (benign ovarian disease). A bivariate mixture model for two serum biomarkers in cases has four components, the first where both biomarkers (STn and CA-125) are overexpressed, the second has STn overexpressed but not CA-125, the third has CA-125 overexpressed but not STn, and the fourth where neither biomarker is overexpressed and hence has the same bivariate distribution as the controls. Each component has a fraction of the cases – with the four fractions summing to 100%. The estimate for a fraction of cases in the second component, where STn is overexpressed but CA-125 is not, estimates the complementarity of STn to CA-125.