Impact of fludarabine and treosulfan on ovarian tumor cells and mesothelin chimeric antigen receptor T cells

In addition to their immunosuppressive effect, cytostatics conditioning prior to adoptive therapy such as chimeric antigen receptor (CAR) T cells may play a role in debulking and remodeling the tumor microenvironment. We investigated in vitro the killing efficacy and impact of treosulfan and fludarabine on ovarian cancer cells expressing mesothelin (MSLN) and effect on MSLN-targeting CAR T cells. Treosulfan and fludarabine had a synergetic effect on killing of SKOV3 and OVCAR4 cells. Sensitivity to the combination of treosulfan and fludarabine was increased when SKOV3 cells expressed MSLN and when OVCAR4 cells were tested in hypoxia, while MSLN cells surface expression by SKOV3 and OVCAR4 cells was not altered after treosulfan or fludarabine exposure. Exposure to treosulfan or fludarabine (10 µM) neither impacted MSLN-CAR T cells degranulation, cytokines production upon challenge with MSLN + OVCAR3 cells, nor induced mitochondrial defects. Combination of treosulfan and fludarabine decreased MSLN-CAR T cells anti-tumor killing in normoxia but not hypoxia. In conclusion, treosulfan and fludarabine killed MSLN + ovarian cancer cells without altering MSLN-CAR T cells functions (at low cytostatics concentration) even in hypoxic conditions, and our data support the use of treosulfan and fludarabine as conditioning drugs prior to MSLN-CAR T cell therapy. Supplementary Information The online version contains supplementary material available at 10.1007/s00262-024-03740-3.


Introduction
Epithelial ovarian cancer is a gynecologic malignancy with very poor prognosis [1].New therapies are urgently needed since, despite extensive surgery and chemotherapy, up to 50% of patients suffer from disease relapse.
Adoptive cell therapy using genetically engineered T cells expressing chimeric antigen receptor (CAR) represent a major breakthrough in cancer treatment but with limited efficacy against solid tumors [2].Most solid tumors contain hypoxic regions representing an inhospitable environment for T cells, contributing to CAR T cell treatments.Mesothelin (MSLN) is a surface glycoprotein with low expression on normal cells and increased expression in many solid tumors including ovarian cancer [3].Several MSLN-associated therapies exist such as MSLN DNA vaccine [4], therapeutic agent biding to MSLN [5], and MSLN-targeting CAR T cells [3].
Cytostatics conditioning prior to CAR T cell therapy plays an important role in hematological malignancies treatment [6] through several mechanisms: by increasing CAR T cell expansion and persistence [7], sensitizing tumor cells to immunotherapy, and inhibiting suppressive immune cells [8].Cyclophosphamide and/or fludarabine (nucleoside analogs) cytostatics are the golden standard conditioning drugs prior to CD19 CAR therapy [9].Treosulfan (alkylating agent) has been previously used in treating ovarian cancer [10] and conditioning prior to hematopoietic stem cell transplantation (HSCT) [11] but not yet tested prior to CAR T cell therapy.Treosulfan and fludarabine combination is commonly used and shown to improve HSCT treatment efficacy and reduce related toxicities [12,13].
In this study, we investigated treosulfan and fludarabine anti-tumor efficacy against ovarian cell lines, impact of MSLN expression and hypoxia, and effect on MSLN-CAR T cells.

Effect of cytostatics on MSLN expression
MSLN + SKOV3 and OVCAR4 (0.15e6) cells were incubated with treosulfan or fludarabine at a concentration corresponding to half of IC 50 (in normoxic conditions, see Table 1), and MSLN expression was measured after 48 h by flow cytometry (CytoFLEX, Beckman Coulter).Untreated cells served as negative control.Results were analyzed using FlowJo Software (BD Life Sciences).

T cell apoptosis assessment
PBMCs (3e5 cells) were incubated for 24 h with treosulfan or fludarabine (10 µM).Cells were stained with Annexin V-APC antibody and 7-AAD in Annexin Binding Buffer (BD Biosciences), and apoptosis was measured by flow cytometric analysis (CytoFLEX).
To measure killing, M28z CAR + T cells were incubated for 24 h with cytostatics, then stimulated for 24 h in normoxia or hypoxia, with MSLN + OVCAR3 cells (effector:target ratio 1:1).Matched untransduced (CAR-) untreated T cells served as negative control.OVCAR3 tumor cells (modified to express luciferase) killing was measured by bioluminescence using One-Glo Luciferase Assay (Promega) and reading with CLARIOstar multireader.M28z CAR T cells-specific killing was calculated as follows:

Data and statistical analysis
Viability of treated cells (WST1 assay) was expressed as percentage of untreated cells viability.Ovarian cancer cells IC 50s was calculated by four-parameter logistic regression model, comparison done using Tukey's multiple comparison test two-way ANOVA, GraphPad Prism (GraphPad software).

Hypoxia and MSLN expression modulate differently the cytotoxicity of treosulfan and fludarabine against SKOV3 and OVCAR4 tumor cells
We evaluated treosulfan and fludarabine IC 50 against SKOV3 and OVCAR4 cells (parental and MSLN+) in normoxia and hypoxia.Treosulfan IC 50 for parental and MSLN + SKOV3 cells were comparable in normoxia and hypoxia (Table 1 and Fig. 1A), and hypoxic incubation induced a significant increase in IC 50 for parental SKOV3 cells (p = 0.005).This impact of hypoxia was not seen for fludarabine IC 50 where a difference (p < 0.0001) was only seen between parental and MSLN + SKOV3 cells in normoxia.When co-incubated with treosulfan, fludarabine IC 50 was significantly lower (p < 0.0001) for MSLN + SKOV3 cells compared to parental cells in normoxia and hypoxia.
With OVCAR4 cells (Table 1 and Fig. 1B), treosulfan IC 50 was significantly increased (p < 0.0001) in hypoxia both for parental and for MSLN + cells.No differences between the different conditions were observed for fludarabine IC 50 , but when co-incubated with treosulfan, fludarabine IC 50 was significantly decreased (p < 0.0001) in hypoxia for parental and MSLN + OVCAR4 cells.
Hypoxia appears to impact the IC 50 of treosulfan and combination of treosulfan and fludarabine on ovarian tumor cells.This impact was cell line-dependent: More differences were observed with (parental and MSLN+) OVCAR4 cells.Hypoxia increased treosulfan IC 50 for SKOV3 and OVCAR4 cells but lowered fludarabine IC 50 when combined with treosulfan for OVCAR4 cells.
MSLN expression modulated SKOV3 cells (but not OVCAR4 cells) drug resistance.When SKOV3 overexpressed MSLN, fludarabine IC 50 was increased in normoxia but decreased when co-incubated with treosulfan in normoxia and hypoxia.
Culturing tumor spheroids with single cytostatics increased their resistance to cytostatics (Fig. 2A, B).Similar trends were observed for SKOV3 and OVCAR4, and the median frequency of live tumor cells (parental and MSLN+, SKOV3, and OVCAR4 cells) retrieved from the spheroids after incubation with treosulfan was between 69-78% and between 60-67% with fludarabine, but was lower (37-58% live tumor cells) when combination of both cytostatics was used.
In addition, MSLN + SKOV3 and OVCAR4 cells incubation with treosulfan and fludarabine for 48 h showed that the frequency (Fig. 2C, D) of MSLN+ expression was unchanged (≥ 95%) compared to preincubation or control.

Treosulfan or fludarabine does not impact MSLN-CAR T cells
We tested the impact of treosulfan or fludarabine on PBMCs viability.Incubation with 10µM treosulfan or fludarabine led to ~ 50% (median, with inter-individual variability) and ~ 12% (median) cell viability after 24 h and 48 h, respectively (Fig. 3A, B).We assessed the impact of 10µM treosulfan and fludarabine on surviving CAR T cells and measured mitochondrial reactive oxygen species (ROS) production, mitochondrial membrane potential, and mitochondrial mass by using MitoSOX, TMRE, and MitoTracker Green dyes, respectively, in different MSLN-CAR T cell products (M28z, MBBz, and M1XX).Incubation for 24 h with treosulfan or fludarabine did not impact the mitochondrial markers regardless of the MSLN-CAR construct or T cell (CD4+ and CD8+) subset.Significant differences were detected in the presence of fludarabine: increased ROS production in CD8 + M28z CAR T cell product and mitochondrial mass of CD4+ and CD8+ in the M1XX CAR T cell product (Fig. 3D).
Similarly, 24-h exposure to treosulfan or fludarabine (10 µM) did not impact MSLN-CAR T cells CD107a expression, cytokine production, or tumor killing when co-incubated with MSLN+ target T cells in normoxia or hypoxia (Fig. 4A-C).However, combined exposure to fludarabine and treosulfan significantly decreased tumor killing in normoxia but not in hypoxia (Fig. 4C).

Discussion
Treosulfan and fludarabine conditioning regimen is widely used for various malignancies.We evaluated in vitro treosulfan and fludarabine (and their combination) as potential conditioning prior to MLSN CAR T cell therapy for ovarian cancer treatment.
We showed that treosulfan and fludarabine mediated the killing and, in combination, had a synergetic effect on SKOV3 and OVCAR4 ovarian cancer cell lines.Since solid tumors form hypoxic niches, we tested the two drugs in normoxic and hypoxic conditions.Hypoxia increased SKOV3 and OVCAR4 resistance to treosulfan, but not fludarabine.Interestingly, when treosulfan and fludarabine were combined, tumor cells drug sensitivity was increased, and a similar trend was observed when using a tumor spheroid model (where lower oxygen levels would be expected).MSLN expression has been shown to promote resistance to some drugs (e.g., platinum and cyclophosphamide combination) [16].MSLN overexpression was previously reported to induce resistance to treatment in pancreatic cancer cells [17].Overexpression of MSLN appeared to modulate sensitivity of SKOV3 cells to fludarabine (increasing it when alone, decreasing when combined with fludarabine), but this was not seen for OVCAR4 cells.Our results highlight the importance of testing relevant tumor cells (and associated antigens of interest) and conditions (e.g., hypoxia) when assessing the cytotoxic effect of cytostatics drugs.
Exposure to treosulfan and fludarabine did not impact MSLN cell surface expression on SKOV3 and OVCAR4 suggesting that these two drugs do not negatively impact this tumor antigen expression and, subsequently, the capacity for MSLN-CAR T cells to bind MLSN and kill tumor cells.
Lymphodepleting conditioning regimens are usually given within a week before CAR T cell treatment [18], and CAR T cells are given within two days after the last cytostatic infusion in order to avoid negative interference.Previous reports have shown that exposure for 24 h with 1µM fludarabine (followed by six days of culture) increased the proliferation and frequency of HDs memory T cells [19].Exposure to treosulfan and fludarabine did not impact MSLN-CAR T cells effectors functions (degranulation and cytokine production) or mitochondrial defects, while combination of treosulfan and fludarabine decreased MSLN-CAR T cells anti-tumor killing in normoxia but not in hypoxia.
Our results indicate that even in combination, these two drugs do not negatively impact MSLN-CAR T cells functions in the context of anhypoxic solid tumor.Altogether, our study shows that treosulfan and fludarabine can mediate cytotoxicity toward SKOV3 and OVCAR4 ovarian cancer cell lines without inducing loss of MSLN cell surface expression or having a negative impact on MSLN-CAR T cells.Further preclinical work is needed to determine the optimal dose and timing for preconditioning with treosulfan and fludarabine prior to MSLN-CAR T cell therapy for the treatment of ovarian cancer.

Fig. 3
Fig. 3 Treosulfan and fludarabine do not impact MSLN-CAR T cells mitochondrial functions.A, B The impact of treosulfan (10 µM) or fludarabine (10 µM) on the viability of healthy donors PBMCs was assessed after 24 h (A) and 48 h (B) by flow cytometry after Annexin V and 7-AAD staining.Blue: early apoptotic, green: dead, and white: live cells (n = 3).C-E Healthy donor MSLN-CAR T cells were exposed for 24 h to treosulfan (10 µM) or fludarabine (10 µM), and