Previously, we and others found that ALDH1A1 and ALDH1A2 mRNAs were upregulated in TIC growth conditions (5) and that ALDH activity is increased in TIC-enriching spheroid growth conditions under the influence of RelB transcriptional activity (17). We selected EOC cell lines with high ALDH activity to test the ALDH1A1 specific inhibitors, NCT-505 and NCT-506. Western blot analysis of EOC cell lines showed that ALDH1A1 protein expression was detected in OV90 and OVCAR3 cell lines, in both adherent and TIC-enriching (spheroid) conditions, but was not detected in OVCAR8, CAOV3, SKOV3 or ACI23 lines ( Figure 1A ). This is consistent with previous reports (20). ALDH1A2 expression was detected only in OV90 cells grown in spheroid conditions ( Figure 1B ). ALDH activity was measured in EOC cells grown in spheroid conditions, and OV90 and OVCAR3 had the highest amount of ALDH activity, followed by ACI23, OVCAR8, and CAOV3 ( Figures 1C, D ). Based on these expression data, we investigated the effects of the specific ALDH1A1 inhibitors on OV90 and OVCAR3 cells. The cell line OVCAR8 was used as a comparator, since its ALDH activity and expression was relatively low, but we previously showed this line was sensitive to the broad ALDH inhibitor disulfiram, suggesting effects beyond ALDH inhibition (15).

The specific inhibitors were compared to disulfiram for their ability to inhibit ALDH activity in OV90 and OVCAR3 spheroids as these lines had the highest ALDH activity of the EOC cells tested. The spheroids were treated with disulfiram or the NCT-505 or -506 compounds for 72 h at the calculated viability IC₅₀ doses for each drug and cell line and activity was assessed by the ALDEFLUOR assay (5, 15, 17). All drugs significantly diminished ALDH activity in spheroids compared to vehicle control but disulfiram did reduce ALDH activity to a greater level than NCT-505 and -506 at the selected concentrations ( Figures 2A, B ).

The specific ALDH1A1 inhibitor compounds NCT-505 and NCT-506 have been reported to reduce viability of OV90 cells grown in TIC-enriching spheroid growth conditions (20). To examine the effect of the specific inhibitors on the viability of EOC spheroid and adherent cells, cell viability was measured at increasing doses of the NCT-505, -506, or disulfiram. OV90 had the strongest expression of ALDH1A1 by Western blot analysis and showed the highest level of ALDH activity of the EOC cell lines tested ( Figure 1 ). In contrast to the effect of disulfiram, the NCT-505 and -506 inhibitors did not achieve the differential effect on viability of OV90 spheroids compared to OV90 cells grown adherently ( Figure 3A ). In OVCAR3 cells, disulfiram and NCT-506 significantly reduced viability of spheroids compared to adherent cells, but NCT-505 did not ( Figure 3B ). Finally, in the low ALDH activity cell line OVCAR8, both NCT-505 and -506 did reduce viability of spheroids compared to adherent ( Figure 3C ), but at a lower potency compared to disulfiram. Disulfiram significantly reduced spheroid viability compared to adherent cell viability in all three cell lines, consistent with previous reports (15) suggesting its broader effects beyond ALDH inhibition and a superior ability to selectively reduce TIC viability.

Sphere formation in EOC cells is important to maintaining TIC survival (21) and inhibition of ALDH by disulfiram has been shown to reduced sphere formation (15). In this study, we also wanted to investigate whether inhibiting ALDH1A1 activity was sufficient to inhibit sphere formation. To test this, cells were grown in the presence of NCT-505, -506 or disulfiram for 7 days and images were acquired to quantify the spheroids that formed ( Figure 4A ). As expected, disulfiram was the most effective in inhibiting sphere formation compared to vehicle control for all lines tested ( Figures 4B–D ). The effect of the ALDH1A1-specific inhibitors on sphere formation was dose-dependent. Using NCT-505 and -506 at the calculated IC₅₀ viability dose ( Figure 4 , “high”), sphere formation efficiency was significantly reduced compared to the vehicle control in OV90 and OVCAR8 ( Figures 4B, D ). However, at the dose that was inhibitory to ALDH activity ( Figure 4 , “low”) the NCTs were ineffective in OVCAR3 and OVCAR8, but did reduce sphere formation in OV90, suggesting higher dependence on ALDH1A1 activity in OV90 compared to the other cell lines ( Figures 4B–D ). Interestingly, at the high dose, NCT-506 had a greater effect than NCT-505 in OVCAR3 and OVCAR8, almost equaling the effect of disulfiram ( Figures 4C, D ).

Disulfiram was an attractive candidate from our previous study as it showed better anti-tumor activity towards EOC spheroids than adherent cells, and was able to prevent relapse as a maintenance drug after carboplatin treatment in vitro and in vivo (15). Therefore we investigated the specific ALDH1A1 inhibitors for their efficacy against spheroid viability after carboplatin treatment, compared to disulfiram in in vitro relapse experiments. OV90 and OVCAR3 cells were treated with carboplatin or vehicle for 48 h in adherent growth conditions, then washed and replated in spheroid conditions in the presence of carboplatin, disulfiram, NCT-505 or -506 for 72 h ( Figure 5A ). OV90 cells tend to be more resistant to platinum than OVCAR3 cells. In OV90 spheroids, NCT-505 and -506 after carboplatin increased cell death compared to vehicle-treated spheroids ( Figure 5B ), but did not extend the cell death effect of carboplatin as a secondary treatment. Disulfiram significantly increased cell death compared to vehicle-treated OV90 spheroids, and compared to spheroids treated with NCT-505 and -506 ( Figure 5B ). In OVCAR3 spheroids, there was a significant amount of cell death where carboplatin was given to adherent cells while the secondary treatment given in spheroid growth conditions did not induce additional cell death ( Figure 5C ), as OVCAR3 cells show greater sensitivity to carboplatin than OV90 cells.

Classical NFκB signaling is elevated in numerous cancers and has been reported to support TIC proliferation (17, 22–24). Reduced NFκB signaling would be a beneficial component of treatment of ovarian TICs, and ultimately recurrent EOC, therefore we examined the effect of each ALDH inhibitor on NFκB activation. Following treatment of OVCAR8 and OV90 TICs with disulfiram at several time points, we measured the expression of p65, which when phosphorylated, signifies the activation of classical NFκB. In OV90 TICs, we observed expression of phosphorylated p65 at the 0, 1, 12, and 18-hour time points, with significantly reduced expression at hours 12 and 18 h compared to hour 0 ( Figure 6A ). In the OVCAR8 TICs, p65 expression patterns differed from that of OV90 TICs after treatment with disulfiram; p65 was observed at all time points, but was significantly reduced at 3 and 12-hour timepoints ( Figure 6B ). OV90 and OVCAR8 TICs were also treated with NCT-505 and -506 and their effects on phosphorylated p65 expression were compared to those of disulfiram at 12 and 24 h. OV90 TICs showed significantly reduced expression of phosphorylated p65 treated with disulfiram and the NCT-505 and -506 ALDH1A1-specific inhibitors ( Figure 6C ). However, OVCAR8 TICs showed increased expression of phosphorylated p65 at 12 and 24 h compared to vehicle following treatment with both NCT-505 and NCT-506 ( Figure 6D ). This suggests that the effect of disulfiram on TICs to reduce NFκB activation depends on the endogenous level of p65 activation. We examined the effect of disulfiram on OVCAR8 viability with inhibition of classical NFκB signaling, using an IKKβ inhibitor, Inhibitor IV. Using a dose that did not significantly affect the viability of OVCAR8 TICs, but did inhibit p65 phosphorylation ( Supplementary Figures 1A, B ), 0.312 µM of Inhibitor IV was added to the OVCAR8 TICs with disulfiram to examine viability. The effect of disulfiram at its previously demonstrated cytotoxic concentrations was attenuated when coupled with Inhibitor IV ( Supplementary Figure 1C ). This finding shows that exogenous NFκB inhibition eliminates the effect of disulfiram on OVCAR8 TICs and suggests that disulfiram confers its cytotoxic effect specifically by inhibiting classical NFκB via the inhibition of p65 phosphorylation.

Carboplatin was purchased from the Tocris Bioscience (Minneapolis, MN) (cat. No. 2626) and dissolved in phosphate buffered saline (PBS). ALDH1A1 inhibitors NCT 505 (NCGC00384406) and NCT 506 (NCGC00386123) were synthesized as described (20) and was provided by Dr. Shyh-Ming Yang (National Center for Advancing Translational Sciences). Propidium Iodide (PI) was from Roche (Cambridge, MA) and Annexin V-FITC (556420) was from BD Biosciences (San Jose, CA). IKKβ inhibitor (Inhibitor IV, 401484) was from Millipore Sigma (Burlington, MA). ALDH1A1 (ab52492) and ALDH1A2 (ab96060) antibodies were from Abcam (Cambridge, MA). GAPDH (MAB374) was from Millipore Sigma (Burlington, MA). Phosphorylated p65 (3033L) cleaved poly (ADP-ribose) polymerase (9541) were from Cell Signaling Technologies (Danvers, MA). Total p65 (10815) was from Santa Cruz Biotechnologies (Dallas, TX).

Ovarian cancer lines OV90, OVCAR3, and OVCAR8 were obtained from the American Type Culture Collection (ATCC, Manassas, VA). All cultures were maintained at 37°C in 5% CO₂. Cell lines were cultured in RPMI (Thermo Fisher Scientific, Waltham, MA) medium containing 10% (v/v) fetal calf serum (FCS), penicillin (100 units per ml) and streptomycin (100 units per ml). TIC-enriching culture medium (TEM) was previously described (17). TIC-enriching spheroid culture conditions were generated by maintaining cells in ultra-low attachment (ULA) plates or flasks (Corning, NY). Experiments involving the TIC-enriched spheroid populations were grown for 3 days in defined medium in ULA plates before drug treatments were performed.

OV90 and OVCAR8 cells were grown in as spheroids for 3 days in TEM, and treated with disulfiram and the NCTs as indicated. Whole cell lysates were collected at the indicated time points post treatment, by pelleting spheroids and removing growth medium and washing in PBS. Lysates were collected in lysis buffer: RIPA buffer (Thermo Scientific, Waltham, MA) containing 1× protease inhibitor (Halt, Thermo Fisher Scientific, Waltham, MA) 1× phosphatase inhibitor (Phos-STOP, Millipore Sigma, Burlington, MA). After a brief incubation on ice, the lysates were homogenized by passing the samples through 26-G needles for 5 strokes, followed by centrifugation at 16,000g, 4°C for 20 min to collect the supernatant. Protein concentration was quantified by microbicinchoninic acid (BCA) assay (Thermo Fisher Scientific, Waltham, MA). Lysates (40 µg) were separated by SDS-PAGE under reducing conditions, transferred onto PVDF membranes, and blocked in 5% non-fat milk in Tris-buffered saline containing 0.1% Tween 20 (TBST). Membranes were incubated with primary antibodies diluted in 1% non-fat milk in TBST overnight at 4°C, washed with TBST, and then incubated with secondary HRP-conjugated mouse or rabbit IgG as appropriate. Images were generated using the Odyssey system and software (LI-COR Biosciences, Lincoln, NE).

Cell viability was assessed as previously described (15) using CellTiter-Glo (Promega, Madison, WI, USA) according to the instructions of the manufacturer.

Sphere formation of OV90, OVCAR8 and OVCAR3 cells was performed as previously described (15). Cells were seeded at 2,000 cells/well in 96-well ULA plates (3474, Corning, Corning, NY), in TEM with indicated drugs for 7 days, fresh culture medium containing growth factors and the drugs was replenished every 48 h. After 7 days the spheres were incubated with DRAQ5 (Thermo Fisher Scientific, Waltham, MA, USA) at 1 µM for 15 min prior to imaging using an inverted Nikon Ti2-E microscope (Nikon, Melville, NY), equipped with a Yokogawa SoRa CSU-W1 spinning disk unit (Yokogawa, Sugar Land, TX) and a BSI sCMOS camera (Teledyne Photometrics, Tuscon, AZ). Images of spheroids were acquired using the automated acquisition module (JOBS) to image each well using a 10× plan-apochromat N.A. 0.45 objective lens, 200 ms exposure. Quantification of spheroids was performed using NIS Elements software version 5.30 (Nikon, Melville, NY). Images were processed using Segment.ai trained on human-recognized, hand-traced spheroids. Segment.ai learned how to trace spheroids in subsequent images. The number of spheroids measuring an area of >1,000 μm² were counted.

ALDH enzymatic activity was quantified using the ALDEFLUOR kit from Stem Cell Technologies (Seattle, WA), used according to the instructions of the manufacturer and as previously described (15). Briefly, cells were plated in spheroid culture conditions for 72 h prior to drug treatment with NCT-505 (11.45 μM), NCT-506 (103 μM), and disulfiram (250 nM). Following 72 h of drug treatment, 1.0 × 10⁵ viable cells were incubated with ALDEFLUOR assay buffer containing the active substrate for 60 min at 37°C or were incubated with an ALDH inhibitor diethylaminobenzaldehyde (DEAB) to serve as a negative control in tandem. Cell death was measured from the in vitro relapse assay, as described previously (15). For the in vitro relapse model, cells were grown in adherent conditions in the presence of carboplatin or PBS control for 48 h, then trypsinized, and 1.0 × 10⁵ cells re-plated in fresh TIC-enriching media in ULA plates and treated with the indicated drugs for 72 h. The cells were collected and a total count was performed, and 1.0 × 10⁵ cells from each treatment were stained with Annexin V-FITC according to the protocol of the manufacturer. After the final wash step, PI was added at 1:10 dilution in PBS and incubated for 15 min, protected from light and directly after was analyzed. Fluorescence was detected on a flow cytometer, and analyzed using FlowJo software (Becton, Dickinson and Company, NJ).

In vitro assays were performed in triplicate on three independent occasions and were analyzed with t-tests or one-way ANOVA with post-tests where applicable. Results are presented as mean ± SEM with p-values ≤0.05 considered significant. Statistical analyses were performed using Prism 8.0 software (GraphPad, San Diego, CA, USA).