Targeting FEN1/EXO1 to enhance efficacy of PARP inhibition in triple-negative breast cancer

Highlights

• •FEN1/EXO1 inhibition enhances PARP inhibition (PARPi) efficacy in 7/10 TNBC cell lines.
• •PARPi and FEN1/EXO1 inhibition synergize in PARPi-resistant cell lines and organoids that are BRCA1/2 wild-type.
• •FEN1/EXO1 inhibition sensitizes cell lines with acquired resistance to PARPi.
• •Synergy is mediated by enhanced DNA damage and DNA replication fork speed.

Abstract

Triple-negative breast cancer (TNBC) is the most aggressive subtype of breast cancer. The only targeted therapeutic approach that has emerged for early TNBC patients with BRCA-mutations (BRCA^MUT) are PARP inhibitors (PARPi). In combination, PARPi may benefit a larger cohort of TNBC patients. We used our previously identified 63-gene signature that was associated with PARPi response to identify candidate genes that could be therapeutic targets. We selected FEN1 for further investigation since its knockdown was associated with an increase in G2/M arrest, DNA damage, and apoptosis. We first tested LNT1, a FEN1/EXO1 inhibitor, in a panel of 10 TNBC cell lines. LNT1 sensitivity was identified predominantly in BRCA1-mutant/deficient cell lines. However, the combination of PARPi and LNT1 demonstrated a synergistic or additive effect in 7/10 cell lines, mainly in BRCA1/2 wild-type (BRCA^WT) and BRCA2-mutant cell lines, with intrinsic and acquired resistance to PARPi. The greatest synergy was observed in a BRCA2-mutant cell line with acquired resistance to olaparib (HCC1395-OlaR), with a combination index value of 0.20. In the synergistic cell lines, BT549 (BRCA^WT) and HCC1395-OlaR, the combination was associated with a rapid progression in DNA replication fork speed, an early and sustained increase in DNA damage in comparison to each of the single-agents. However, in the additive BRCA1/2 wild-type cell lines, MDAMB231 and HCC1806, the combination demonstrated a high DNA damage response that was largely driven by either talazoparib or LNT1. Therefore, targeting FEN1/EXO1 with PARPi is a promising targeted combination approach, particularly in the context of PARPi-resistant and BRCA^WT TNBC.

Graphical abstract

Introduction

Triple-negative breast cancer (TNBC) is the most aggressive subtype of breast cancer. TNBC comprises 15–20 % of all breast cancers and has limited treatment options, as these tumors lack overexpression of estrogen and progesterone receptors (ER/PR) and human epidermal growth factor receptor 2 (HER2) [1]. Among TNBC patients, 11–20 % carry germline mutations in the BRCA1/2 genes (gBRCA^MUT), which has been associated with an increase in breast cancer risk [2]. In cells, BRCA1/2 function in the homologous recombination repair pathway to mend double strand breaks (DSBs) in DNA [3]. Loss of function of BRCA1/2, as in gBRCA^MUT cases, leads to improper repair of DSBs and contributes to the development of breast cancer.

Breast cancer patients who are gBRCA^MUT benefit from Poly (ADP-ribose) polymerase (PARP) inhibitors (PARPi), such as talazoparib and olaparib, which have been shown to improve survival [1]. PARPi act by synthetic lethality and PARP-DNA trapping, with talazoparib being the most potent in PARP-DNA trapping [4]. However, studies have shown that PARPi may also be effective in tumors that actually lack the BRCA1/2-mutation, yet demonstrate ‘BRCAness’, with a functional defect in DNA repair commonly attributed to impaired homologous recombination [5,6]. Several biomarkers of BRCAness have been tested in patient samples including those derived from whole genome sequencing (HRDetect), genomic scar assays, gene expression signatures, and mutational signatures [1].

We previously derived a 63-gene signature predictive of PARPi response in TNBC [7]. Using the DNA damage response of three PARPi in a panel of TNBC cell lines, we showed that PARPi response was associated with an enrichment of pathways including DNA double strand break repair, DNA damage bypass, base excision repair, and cell cycle checkpoints. Our gene signature demonstrated an overall accuracy of 86 % in a small cohort of patient-derived xenografts treated with olaparib, and predicted sensitivity in 45 % of TNBC patients, which was comparable to other biomarkers of BRCAness [1]. However, in addition to being a predictive biomarker, our gene signature can also be used to identify therapeutic targets [7]. These targets can enhance the activity of anti-PARP therapy by targeting pathways involved in DNA repair or cell cycle checkpoints.

Clinically promising therapeutic targets have emerged from our 63-gene signature. For example, CHK1, involved in checkpoint mediated cell cycle arrest, and USP1 (ubiquitin specific peptidase 1), involved in DNA damage bypass, have demonstrated efficacy when targeted in combination with PARPi preclinically [8,9], and are being tested in Phase 1 clinical trials in solid tumors [10,11]. PARPi have also been combined with inhibitors of ATM, a DNA damage kinase, or WEE1, a regulator of cyclin-dependent kinase 1/2 [[12], [13], [14]]. Hence, there is precedence to combine PARPi with drugs that target DNA repair or cell cycle checkpoints [15].

Here, we screened targets from the 63-gene signature and selected Flap Endonuclease 1 (FEN1) as a target for PARPi combination therapy. FEN1 is a replication-regulating protein, involved in base excision repair, and can be recruited to single-strand DNA damage sites and Okazaki sites by active PARP-1 [16]. Interestingly, FEN1 has been identified in other PARPi sensitivity screens [17,18]. Here, we used the FEN1/EXO1 inhibitor LNT1 [19] in chemosensitivity assays to test its efficacy in combination with the PARPi talazoparib in a panel of TNBC cell line and xenograft-derived organoid (XDO) models. We show that talazoparib is synergistic with LNT1 in TNBC cell lines with intrinsic and acquired resistance to PARPi. These cell lines demonstrated a marked increase in DNA fork replication speed and DNA damage. Taken together, our results suggest that targeting FEN1/EXO1 is a promising combination approach with anti-PARP therapy.

Materials and methods

Cell culture

A complete list of cell lines, sources, and corresponding media is available in Supplementary Table S1. Cell lines were validated by DNA fingerprinting using short-tandem repeat (STR) analysis, last performed in April 2024. Cell lines and organoids were tested regularly to confirm the absence of mycoplasma using Mycoalert mycoplasma detection kit (LT07, Lonza).

Acquired resistance cell line derivation

PARPi-resistant HCC1395 cells were derived from HCC1395 TNBC cells grown in RPMI with increasing concentrations of talazoparib or olaparib added to the media over the course of six months, as previously described [20]. Cells were passaged in T-25 flasks and allowed to adhere. The next day, media containing 0.0006 µM talazoparib or 0.14 µM olaparib was added. Cells were grown in the presence of either drug until confluence was reduced to approximately 10 %, then allowed to recover to near confluence in complete media lacking PARPi. Media was exchanged every 3–5 days, alternating between media containing or lacking PARPi depending on cell growth. Treatment with PARPi was repeated at the same concentration until resistance was observed, at which point PARPi concentration was doubled. This process was repeated until sufficient resistance was observed. Finally, stable resistance was confirmed by IC₅₀ calculation after 5 passages in the absence of PARPi and a freeze-thaw cycle.

Western blots

Cells were lysed in a cell lysis buffer (Cat. #9803, Cell Signaling Technology) containing phosphatase inhibitors (Cat. #4,906,845,001, Roche) and PMSF (Cat. #8553S, Cell Signaling Technology), followed by centrifugation at 14,000 x g for 10 min at 4 °C. Protein concentrations were quantified using the Pierce™ BSA Protein Assay (Cat. #23,227, Thermofisher). Equal amounts of protein were run on 8–12 % acrylamide gels at 90 V for 30 min and 120 V for 1–2 h, then transferred to nitrocellulose membranes for 1 hour at 100 V. Membranes were blocked with 5 % BSA or milk in TBST buffer for 1 hour at room temperature, then incubated with primary antibodies overnight at 4 °C. Membranes were washed 3 × 10 min in TBST buffer and incubated with the appropriate HRP-conjugated secondary antibodies for 1–2 h at room temperature. Finally, membranes were washed 3 × 5 min in TBST, stored in TBS, and exposed to Clarity ECL (Biorad) or PICO ECL supersignal (ThermoFisher) to be imaged with the ChemiDoc Imaging System. All antibodies are listed in Supplementary Table S2.

Xenograft-derived organoids

Organoids were developed from mice tumors obtained from orthotopic xenograft experiments that we previously performed [21]. All animal experiments were approved by the Institutional Animal Protection Committee (CIPA) of the Centre de Recherche de Centre hospitalier de l'Université de Montréal (CRCHUM) under protocol C17017SHs. Two million HCC1806 cells were surgically implanted in the mammary fat pad of NOD-SCID gamma (NSG) female mice (Cat. #005557, Jackson Laboratory). Once untreated tumors reached a mean volume of 12 mm³, mice were sacrificed and tumors harvested.

Fresh tumor samples were processed and organoid models established as described [22].

XDO maintenance media was adapted from Type 1 medium [22] lacking primocin, with the following adjustments: 100 ng/mL R-spondin 1 (Cat. #120–38–100UG, PeproTech), 100 ng/mL Noggin (Cat. #120–10C-100UG, PeproTech), 5 mM nicotinamide (Cat. #A15970.22, Thermo Scientific), and 500 nM SB202190 (Cat. #202190, StemCell Technologies). For 10-day chemosensitivity assays, 96-well plates (Cat. #3610, Corning) were pre-coated with 10 µL of Cultrex Reduced Growth Factor Basement Membrane Extract (Cat. #3433–005–01, R&D Biosystems). XDOs were trypsinized to single cells, resuspended at 20 000 cells/mL, and 100 µL of cell suspension was plated per well. Media lacking N-acetylcysteine and Y-27632 was used for XDO chemosensitivity assays, as previously performed [23]. XDOs were treated with talazoparib, LNT1, or the combination as described above, for a 10-day assay. Cell viability was measured by CellTiter-Glo 3D Cell Viability Assay (Cat. #G9682, Promega) according to manufacturer protocols.

RNA sequencing (RNA-seq) and analysis

RNA sequencing was performed on 3 independent samples collected from each cell line. RNA was extracted using RNeasy Plus mini kit (Cat. #74,134). RNA was quantified using Qubit (Thermo Scientific) and quality was assessed with the 2100 Bioanalyzer (Agilent Technologies). Libraries were prepared and samples were sequenced with Illumina NextSeq 500, using the Illumina kit NextSeq 75 cycles High Output v2, with ∼ 25 M reads per sample, at the Institute for Research in Immunology and Cancer (IRIC, Montreal).

The first part of the analysis including normalization and DeSEq2 analysis was performed at the Bioinformatics Facility at IRIC. Sequences were trimmed for sequencing adapters and low quality 3′ bases using Trimmomatic version 0.35 [24] and aligned to the reference human genome version GRCh38 (gene annotation from Gencode version 37, based on Ensembl 103) using STAR version 2.7.1a [25]. DESeq2 version 1.30.1 [26] was then used to normalize gene read counts and produce the sample clustering. Using the log2-fold change values from the DESeq2 analysis, all expressed genes were ranked and a pre-ranked Gene Set Enrichment Analysis was performed using GenePattern [27]. Gene sets from MsigDB were used, including Hallmarks and KEGG legacy for human tissue. FDR cutoff of 25 % and p-value < 0.05 were used to identify statistically significant gene sets. Dot plots and heatmaps of hierarchical clustering were created in RStudio v4.2.2 [28].

Statistical analysis

Graphs with statistical analysis were made in GraphPad Prism 10, with the exception of an online tool, gene expression-based outcome for breast cancer, for clinical correlations [29]. Experiments were performed with a minimum of n = 2 independent replicates, with 3 replicate wells each for experiments performed in 96-well plates. Significance was calculated by one- or two-way ANOVA with Tukey HSD as appropriate, apart from DNA fiber assays, which were analysed by Kruskal-Wallis ANOVA with multiple comparisons. P < 0.05 is considered statistically significant.

Further details regarding siRNA transfections, chemosensitivity assays for IC₅₀ and combination index values, immunofluorescence, high-content imaging, DNA fiber assays, and flow cytometry are provided in Supplementary Methods [7,21].

Results

Knockdown of selected genes from 63-gene signature increases cellular response to talazoparib

We previously derived a 63-gene signature of PARPi response from a panel of TNBC cell lines treated with veliparib, olaparib, or talazoparib [7]. We selected talazoparib for this study since this is a potent PARPi [4], with similar clinical efficacy as olaparib [1]. From the signature, we identified seven genes of interest: BRCA1, BRCA1-associated RING domain protein 1 (BARD1), Mitotic checkpoint serine/threonine-protein kinase (BUB1), Ribonucleotide reductase regulatory subunit M2 (RRM2), USP1, FEN1, and Exonuclease 1 (EXO1). Targeted gene silencing using siRNA knockdown was performed in a BRCA1/2-wild type (BRCA^WT) cell line, MDAMB231, which was known to overexpress each of these proteins (Supplementary Fig. S1A). To determine whether targeting these proteins would increase the efficacy of talazoparib treatment, we used flow cytometry to quantify cell cycle changes, the proportion of γ-H2AX-positive (γ-H2AX+) cells, a marker of DSBs, and cleaved-Caspase 3-positive (cl-Caspase 3+) cells, a marker of apoptosis (Supplementary Fig. S1B-E).

We focused on the gene knockdowns that demonstrated a similar impact on cell cycle changes, DNA damage, and an enhanced effect in combination with talazoparib in comparison to its respective gene knockdown alone. The impact of siBARD1 was very similar to what was observed with siBRCA1 with or without talazoparib, with a similar increase in cells in G2/M arrest, γ-H2AX+cells, and cl-Caspase 3+ cells in comparison to talazoparib alone or control. These findings are to be expected considering that BARD1 is an obligatory partner of BRCA1, involved in homologous recombination and DNA replication fork stability [30]. In comparison to single-gene knockdown, the combination of talazoparib with either siBUB1 siUSP1, siFEN1, or siEXO1 resulted in similar increases in the percentage of cells in G2/M arrest (7.4 %−9.7 %), and similar increases in γ-H2AX+ cells (19.6 % to 26.9 %). However, despite the similar range of induction of DSBs, silencing of different genes led to a differential impact upon apoptosis, with siFEN1 plus talazoparib demonstrating the highest increase in cl-Caspase 3+ cells by 17.0 % and 18.0 % (P < 0.0001; P < 0.0001), in comparison to siFEN1 and talazoparib, respectively, suggesting that cell death may be occurring by DNA damage-dependent and independent mechanisms. Furthermore, the siFEN1 plus talazoparib combination resulted in the highest levels of cl-Caspase 3+ cells in comparison to the other three combinations.

To better understand the clinical significance of these targets, we evaluated the expression of these genes in different breast cancer subtypes in a cohort of 881 untreated breast cancer patients [29]. We identified elevated expression of USP1, RRM2, FEN1, and EXO1 in more aggressive breast cancer subtypes (basal and HER2), and that their overexpression were associated with a poorer 10-year distant metastasis-free survival (Supplementary Fig. S1F-I). Of note, no statistical significance in distant metastasis-free survival was identified for BARD1 or BUB1 (data not shown).

To summarize, our 63-gene signature was used to identify genes that can be targeted and enhance sensitivity to talazoparib. Since the greatest induction of apoptosis was demonstrated with siFEN1 plus talazoparib in comparison to control or talazoparib, and because FEN1 demonstrated strong prognostic value, we selected FEN1 for further investigation. We used LNT1, a commercially available chemical inhibitor of FEN1 [31]. However, since LNT1 inhibits FEN1 and EXO1 with equal potency and specificity [19], LNT1 will be referred to as a FEN1/EXO1 inhibitor from here onwards.

LNT1 is synergistic with talazoparib in PARPi-resistant cell lines

First, to compare the impact of LNT1 with siFEN1 and siEXO1, we evaluated cell cycle changes in MDAMB231 (Supplementary Fig. S2). In comparison to control, the combination of LNT1 plus talazoparib increased the percentage of cells in G2/M arrest by 11 %, which was similar to the proportion of cells in G2/M arrest induced by siFEN1/siEXO1 and talazoparib, at 11.8 % and 9.4 %, respectively (Supplementary Fig. S1C).

Using a 10-day chemosensitivity assay, we determined IC₅₀ values for LNT1 and talazoparib in TNBC cell lines (Fig. 1A). We used a panel of TNBC cell lines whose molecular features are summarized in Fig. 1B [21]. We developed two cell lines with acquired resistance to PARPi by exposing the PARPi-sensitive HCC1395 TNBC cell line (BRCA2-mutant) to increasing concentrations of olaparib (olaparib-resistant, OlaR) or talazoparib (talazoparib-resistant, TalaR) (Fig. 1C). These cell lines reached resistance to talazoparib in a similar manner to our most intrinsically resistant cell line, BT549 (BRCA^WT) (Fig. 1D). The IC₅₀ values for talazoparib among our cell lines ranged from 3.0 × 10⁻⁴ µM to 0.47 µM, while LNT1 sensitivity was in a much smaller range, from 1.4 µM to 15.7 µM (Fig. 1D, E and Supplementary Fig. S3). LNT1 was most effective in SUM149PT (BRCA1-mutant), MX1 (BRCA1-deleted and BRCA2-mutant), MDAMB436 (BRCA1-mutant), and HCC1806 (BRCA^WT), cell lines which are also sensitive to talazoparib. While HCC1395 demonstrated sensitivity to talazoparib, the cell line was resistant to LNT1. However, HCC1395-OlaR, HCC1395-TalaR, and Hs578T were similarly resistant to each of the compounds. Therefore, LNT1 was most sensitive in BRCA1-mutant and BRCA^WT cell lines.

Fig. 1.

LNT1 enhances sensitivity to PARPi in most TNBC cell lines. (A) Schematic representation of the 10-day chemosensitivity assay, wherein cells are seeded in a 96-well plate, followed by drug administration 24 h later, and cells fixed and stained at 10 days. Each cell line is treated with 9 concentrations of each drug for the IC₅₀ curve derivation. For CI values, 6 concentrations of each drug, either alone or in combination were used (Supplementary Table S3). (B) Molecular characteristics for each cell line. (C) HCC1395 cell lines were derived with acquired resistance to PARP inhibitors. IC₅₀ dose-response curves for HCC1395-Parental (HCC1395-Par) (black), HCC1395-Olaparib Resistant (OlaR) (orange), and HCC1395-Talazoparib Resistant (TalaR) (blue) cell lines. IC₅₀ values for the resistant cell lines and ratio relative to HCC1395-Parental are indicated in table below. IC₅₀ values for (D) talazoparib, and (E) LNT1 in TNBC cell lines. Images of DAPI-stained cells used for cell counts for calculating IC50 curves from representative cell lines are shown in Supplementary Fig. S3. Data represent mean + /− SEM, with individual replicates indicated. (F) Combination index values for talazoparib and LNT1 reported at an FA of 0.5. Teal green bars/dots represent BRCA^WT cell lines, and pink bars/dots represent BRCA^MUT cell lines.

Next, we determined the efficacy of the combination of LNT1 and talazoparib by calculating Combination Index (CI) values. Using the same 10-day chemosensitivity assay, cells were treated with six concentrations of each drug alone or in combination (Supplementary Table S3). Here, we considered CI < 0.9 to be synergistic, 0.9 < CI < 1.1 as additive, and CI > 1.1 as antagonistic [32]. Overall, four TNBC cell lines were synergistic, three were additive, and three were effectively antagonistic (Fig. 1D). SUM149PT was borderline antagonistic, with a CI value of 1.1, while Hs578T and MDAMB436 had CI values of 1.2 and 1.4, respectively. Surprisingly, the cell lines that were highly resistant to talazoparib were most synergistic with LNT1, including BT549 (0.7), HCC1395-OlaR (0.2), and HCC1395-TalaR (0.7). Among these three cell lines, the dose reduction index of talazoparib ranged from 3.4 to 14.7 and from 1.8 to 7.5 for LNT1. Hence, to achieve synergy in these PARPi-resistant cell lines, the concentrations of LNT1 and talazoparib required can be markedly reduced.

Talazoparib and LNT1 induce DNA damage and apoptosis in PARPi-resistant cell lines

To understand the impact of the combination of talazoparib and LNT1 on early DNA damage, we evaluated percentage of γ-H2AX+ cells using immunofluorescence, from 30 min to 72 h post-treatment in BT549, HCC1395-OlaR, and MDAMB231 cell lines (Fig. 2A–E and Supplementary Fig. S4). All three cell lines demonstrated greater percentage of γ-H2AX+ cells with the combination in comparison to control at 24 h, 48 h, and 72 h post-treatment. However, in the more synergistic cell lines, BT549, HCC1395-OlaR, talazoparib alone induced less DNA damage, which is in contrast to the additive cell line, MDAMB231. At 24 h, in the synergistic cell lines, the combination resulted in a 1.8-fold greater induction in γ-H2AX+ cells versus talazoparib in comparison to the 1.1-fold difference in the additive cell line. At 72 h, we identified similar induction of γ-H2AX expression with the combination in comparison to control in BT549 cells by immunoblots (Fig. 2D). Moreover, p-p53 was overexpressed with talazoparib alone and the combination (Fig. 2D). Since p53 may trigger apoptosis, our results suggest that talazoparib alone may induce apoptosis in a DNA-damage independent manner, which may be further enhanced in combination with LNT1.

Fig. 2.

Early induction of DNA damage with the combination of PARPi and LNT1. Using immunofluorescence, percentage of positive γ-H2AX+ cells were calculated for cells treated with DMSO control, talazoparib, LNT1 or the combination of talazoparib and LNT1, for 30 min, 24 h, 48 h, and 72 h in (A) BT549 (B) HCC1395-OlaR and (C) MDAMB231 cell lines. (D) Immunoblots were used to determine γ-H2AX and p-p53 expression in BT549 cell after 72 h treatment of either control (C), talazoparib (T), LNT1 (L), or talazoparib plus LNT1 (TL). Bar graphs of mean percentage of positive γ-H2AX+ cells/γ-H2AX expression/p-p53, where error bars refer standard error of the mean. Two-way ANOVA was performed to test statistical significance within each time point, with P < 0.0001****, P < 0.001***, P < 0.01**, P < 0.05*. (E) Representative 40x objective images of BT549, HCC1395-OlaR, and MDAMB231 cells treated with control (far left), talazoparib (left), LNT1 (right), or the combination (far right). Cells were stained for γ-H2AX (red), with HCS nuclear mask (blue). Scale bars indicate 25 µm.

To determine the impact of talazoparib and LNT1 on sustained DNA damage and apoptosis by immunofluorescence, we used the 10-day chemosensitivity assay. Cells were treated with increasing concentrations of LNT1 (L1-L6), talazoparib (T1-T6), or the combination (TL1-TL6), then fixed and stained for 53BP1 and cleaved-PARP. We calculated a 53BP1 product score for DNA damage using the product of the mean number of nuclear 53BP1 foci per cell and the percentage of cells positive for 53BP1 (Fig. 3 and Supplementary Figs. S5,S6,S7) [21]. In 5/6 cell lines that we tested, the combination of LNT1 with talazoparib induced DNA damage at 2–4-fold lower concentrations than either agent alone (Fig. 3A). In HCC1395-OlaR cells, the most synergistic cell line, we observed DNA damage at 4-fold dose reduction of both talazoparib and LNT1. We also evaluated the 53BP1 product score at specific drug concentrations for each cell line (Fig. 3C and Supplementary Fig. S5). The 2-drug combination demonstrated a 10 and 21-fold increase in 53BP1 product score in HCC1395-OlaR cells, in comparison to talazoparib and LNT1 respectively. However, in the two cell lines which demonstrated an additive effect of the combination, HCC1806 and MDAMB231, a more modest 2–3-fold increase in the 53BP1 product score was observed with the combination in comparison to either of the single agents.

Fig. 3.

Combined talazoparib and LNT1 induces DNA damage at reduced concentrations in synergistic cell lines. Using the 10-day chemosensitivity assay as described in Fig. 1A, 53BP1 foci and cleaved-PARP expression was analyzed. Heatmaps of normalized (A) 53BP1 product score (product of mean 53BP1 foci per nuclei and percentage of cells positive for 53BP1) and (B) percentage of cleaved-PARP positive (cl-PARP+) cells. The x-axis indicates increasing concentrations (1–6) of talazoparib (T or Tala), LNT1 (L), or the combination (TL), with concentrations centered around the IC₅₀ for each cell line. (C) Quantification of 53BP1 product score (top panel) and cl-PARP+ cells (bottom panel) at selected concentrations of talazoparib, LNT1, or the combination of talazoparib and LNT1. Bar graphs of mean 53BP1 product score, where error bars refer to standard error of the mean. Two-way ANOVA was performed to test statistical significance, with P < 0.0001****, P < 0.001***, P < 0.01**, P < 0.05*. (D) Representative 20x objective images of HCC1395-OlaR cells treated with talazoparib (left), LNT1 (middle), or the combination (right). Cells were stained for 53BP1 (red), cl-PARP (green), and nuclear staining with HCS nuclear mask (blue). Scale bars indicate 25 µm. Single channel and merged images of HCC1395-OlaR and additional cell lines are shown in Supplementary Figs. S6 and S7.

Apoptosis was measured by the percentage of cells positive for cleaved-PARP (cl-PARP+) Fig. 3B) [21]. Similar to DNA damage, cell death was most pronounced in the HCC1395-treated cell lines. Indeed, in the synergistic HCC1395-parental, HCC1395-OlaR, and HCC1395-TalaR cell lines, dose reductions for the combination was noteworthy, ranging from 2 to 8-fold. The overall induction of cl-PARP+ cells was low at the concentrations used, suggesting that there may be other mechanisms of cell death. Of note, HCC1395-OlaR cells exhibited high levels of cl-PARP+ cells even at low concentrations of talazoparib alone. This may indicate that HCC1395-OlaR cells are adapted to endure a high level of DNA damage, as induced by talazoparib treatment. Interestingly, the most synergistic cell line, HCC1395-OlaR demonstrated an 8-fold higher percentage of cl-PARP+ cells with the combination in comparison to LNT1 alone (Fig. 3C).

Combined PARPi and LNT1 increases DNA replication fork speed

PARPi or FEN1 depletion have previously been shown to increase replication fork speed [21,33]. Therefore, we investigated the impact of the combination of 72 h treatments with talazoparib and/or LNT1 on fork replication speed in four TNBC cell lines (Fig. 4 and Supplementary Fig. S8). We observed differences in endogenous replication speed depending on the cell line, but consistent increases in fork speed in talazoparib treated cells. In contrast, LNT1 increased the fork speed by 26 % (P < 0.0001) in only one cell line BT549, a synergistic cell line (Fig. 4A). Interestingly, the highest increase in fork speed induced by the combination was also demonstrated in BT549 in comparison to control (47.4 %, P < 0.0001), and ∼17 % increase in comparison to either talazoparib alone (P = 0.008) or LNT1 alone (P = 0.0002). In the most synergistic cell line, HCC1395-OlaR, the combination increased fork speed by 30.9 % in comparison to control (P < 0.0001) and 26.1 % in comparison to talazoparib alone (P < 0.0001) (Fig. 4B). In MDAMB231, which demonstrated a nearly additive effect of the combination, fork speed increased by 22.5 % in comparison to control (P < 0.0001) (Fig. 4C). Finally, another cell line which also showed a nearly additive effect with the combination, MX1, revealed an increase in fork speed by 22.1 % in comparison to control (P < 0.0001) and a 10.0 % increase in comparison to talazoparib (P = 0.002) (Fig. 4D). Overall, since high fork speeds were more pronounced in the synergistic cell lines, acceleration of fork speed may be an important mechanism explaining the induction of DNA damage and efficacy of the combination of talazoparib and LNT1.

Fig. 4.

Talazoparib and LNT1 increase DNA replication fork speed in TNBC cells. After 72 h of drug treatment, the DNA fiber assay was performed. Herein, cells were sequentially labelled with CIdu and IdU, DNA denatured, and then stained for BrdU. (A-D) DNA fork speed for (A) BT549, (B) HCC1395-OlaR, (C) MDAMB231, and (D) MX1 cells measured by DNA fiber assay, with a mean count of 605 fibres per condition. Dot plots of fork speed (kb/min). Lines indicate median values with interquartile range. Kruskal-Wallis test was performed in GraphPad Prism. P < 0.0001****, P < 0.001***, P < 0.01**. (E) Representative images of HCC1395-OlaR DNA fibres at 63x objective. DNA was labelled with CldU (magenta) and IdU (green) thymidine analogues. Scale bars represent 10 µm.

Combined PARPi and LNT1 are synergistic in a xenograft-derived organoid model of TNBC

To evaluate the potential of talazoparib and LNT1 in a more stringent model, we established xenograft-derived organoids (XDOs) (Fig. 5A). Here, we developed organoids from primary tumors in mice that were implanted with the HCC1806 cell line. Once XDOs were established, we characterized the XDOs for their sensitivity to both talazoparib and LNT1 treatment by performing IC₅₀ experiments (Fig. 5B and C). We found that, relative to the IC₅₀ values we measured across cell lines (Fig. 1A and B), HCC1806 XDOs were more resistant to talazoparib, with an IC50 of 0.11 µM, and slightly resistant to LNT1, with an IC50 of 12.5 µM (Fig. 5C). Thus, based on this high resistance to talazoparib, we considered that HCC1806 XDOs, in contrast to the cell line counterpart, might exhibit synergism between talazoparib and LNT1. Indeed, when treated with the combination, we calculated the CI of talazoparib with LNT1 as 0.80 for HCC1806 XDOs (Fig. 5D). Thus, similar to what we observed in cell lines, we observed that LNT1 synergizes with talazoparib in talazoparib-resistant contexts.

Fig. 5.

Talazoparib and LNT1 are synergistic in xenograft-derived organoid (XDO) model of TNBC. (A) Schematic representation of generation of XDO. Cells from an established TNBC cell line, HCC1806, were implanted into the mammary fat pad of mice from which tumors were grown and harvested for organoid development. Tumors were dissociated to single cells and embedded in basement membrane extract for growth. (B) Representative IC₅₀ dose-response curves of talazoparib (top) and LNT1 (bottom) in HCC1806 XDOs. Dose-response curves include nine concentrations of talazoparib from 0.013 nM to 5 μM with 1/5 dilutions, and 9 concentrations of LNT1 from 0.39 µM to 100 μM with 1/2 dilutions. (C) Representative images of HCC1806 XDO cells treated with talazoparib (top), LNT1 (middle), or the combination (bottom). IC₅₀ values for talazoparib (Tala) and LNT1, and CI (Combination Index) for the combination are indicated below.

Genetic determinants of response to LNT1 and synergy with talazoparib

To better understand the role of FEN1, EXO1, PARP-1, and PARG expression in influencing response to LNT1, we evaluated the RNA expression of each of these genes across our panel of cell lines (Fig. 6A, B and Supplementary Fig. S9). We did not find any correlation with FEN1 RNA or protein expression with LNT1 response (Supplementary Fig. S9A). While EXO1 expression varied across cell lines, we also decided to evaluate the EXO1:FEN1 ratio as they are the dual targets of LNT1 (Fig. 6C). Interestingly, most of the LNT1-sensitive cell lines, which are mainly BRCA1-mutant, demonstrated a high EXO1:FEN1 ratio, whereas all the LNT1-resistant cell lines demonstrated low EXO1:FEN1 ratios. HCC1395-parental, the BRCA2-mutant cell line, demonstrated high PARP-1 and FEN1 expression. Since in the context of BRCA2-deficient tumors that are PARPi-resistant, loss of Poly (ADP-Ribose) glycohydrolase (PARG) was synthetically lethal with FEN1 and EXO1 [34], we also evaluated PARG RNA expression (Supplementary Fig. S9F). We did not observe decreased expression of PARG in the HCC1395-OlaR/TalaR cell lines.

Fig. 6.

Evaluating intrinsic gene expression associated with treatment response. Using DeSeq2 normalized RNA-seq gene expression values across the panel of TNBC cell lines to evaluate (A) FEN1, (B) EXO1, and (C) ratio of EXO1:FEN1 expression. (D) Pre-ranked Gene Set Enrichment Analysis (GSEA) of cell lines comparing PARPi-resistant cell lines that became synergistic (HCC1395-OlaR, HCC1395-TalaR, and BT549) versus PARPi-resistant cell line that remained resistant (Hs578T) with dot plot summarizing the most statistically significant enriched pathways. Pathway names are indicated on the y-axis. Colour intensity indicates the false discovery rate (FDR) and dot size indicates the number of genes in each pathway. The x-axis represents nominal enrichment score (NES), with negative values indicating downregulated pathways and positive values indicating upregulated pathways. (E) Representative enrichment plots of the most upregulated (top, ribosome) and most downregulated (bottom, epithelial-mesenchymal-transition) pathways based on NES. (F) Pre-ranked GSEA comparing synergistic or additive cell lines (BT549, HCC1395-OlaR, HCC1395-TalaR, HCC1395 parental, MDAMB231, HCC1806) in comparison to antagonistic cell lines (Hs578T, SUM149PT, MDAMB436). Table summarizes statistically significant pathways that are upregulated.

To investigate the mechanism driving acquired resistance to PARPi in HCC1395 cells, we used a pre-ranked GSEA to compare HCC1395-OlaR and HCC1395-TalaR cells versus HCC1395-Parental. The most upregulated pathway in HCC1395-OlaR/TalaR was ribosome (NES 2.05; FDR 0.004) (Supplementary Fig. S10). We also correlated the expression of four genes that were differentially expressed between the OlaRes/TalaRes versus HCC1395 parental cell lines using RNAseq and rt-PCR (Supplementary Fig. S11). We then wanted to better understand which features of the PARPi-resistant cell lines might be modulated by LNT1, making them sensitive to PARPi. To do so, we performed a pre-ranked Gene Set Enrichment Analysis (GSEA), comparing RNA expression of untreated PARPi-resistant cell lines that became sensitive with LNT1 (BT549, HCC1395-OlaR, HCC1395-TalaR) versus a PARPi resistant cell that remained resistant despite LNT1 treatment (Hs578T) (Fig. 6D). Upregulated pathways include ribosome (Nominal Enrichment Score (NES) 2.18; False Discovery Rate (FDR) <0.001), Myc targets (NES 1.55; FDR 0.10), mismatch repair (NES 1.53; FDR 0.10), UV response up (NES 1.45; FDR 0.13), and chemokine signalling (NES 1.31; FDR 0.23) (Fig. 6E). Downregulated pathways include epithelial-mesenchymal transition (NES −2.32; FDR <0.001) and apoptosis (NES −1.52; FDR 0.11). This is suggestive that in addition to mismatch repair and DNA damage response genes, FEN1/EXO1 inhibition is probably implicated in downregulating ribosomal and chemokine signalling genes, which in turn increases sensitivity to PARPi. This is also indicative that the upregulation of ribosomal genes may be an important mechanism driving intrinsic and acquired resistance to PARPi.

Alternatively, to better understand which features of the LNT1-resistant cell lines might be modulated by PARPi, making them sensitive to LNT1, a pre-ranked GSEA was performed, comparing RNA expression of untreated LNT1-resistant cell lines that became sensitive with PARPi (HCC1395-Parental, HCC1395-OlaR, HCC1395-TalaR, MDAMB231) versus a LNT1-resistant cell line that remained resistant despite PARPi (Hs578T). The downregulated and upregulated pathways were very similar to the analysis performed above, with upregulated pathways including mismatch repair (NES 1.64; FDR 0.05, upregulated genes include EXO1 and RFC3), and base excision repair (NES 1.43; FDR 0.14, upregulated genes include POLE2, UNG, PARP-4, PARP-1, and PARP-2) (data not shown). Finally, to better understand which pathways are involved in the antagonistic response, we performed a pre-ranked GSEA to compare the synergistic or additive cell lines (BT549, HCC1395-parental, HCC1395-OlaR, HCC1395-TalaR, HCC1806, MDAMB231, and MX1) versus antagonistic cell lines (Hs578T, SUM149PT, and MDAMB436). We identified upregulated pathways including mismatch repair, DNA replication, E2F Targets, G2/M checkpoint (Fig. 6F), suggesting that competencies in cell cycle progression may be associated with resistance to the combination.

Discussion

With the increasing utility of PARPi in the clinic [35], improved strategies are required to tackle resistance to PARPi. We identified several promising candidate genes from our 63-gene signature that can be targeted in combination with PARPi. We found that amongst USP1, FEN1, and EXO1, FEN1 knockdown plus PARPi induced the most apoptosis, thereby providing rationale for selecting FEN1 for further investigation. To our knowledge, we are the first to report the impact of the combination of FEN1/EXO1 inhibition and PARPi in a panel of patient-derived BRCA^MUT and BRCA^WT TNBC cell line and organoid models. We showed that the combination of LNT1 and talazoparib demonstrated an additive or synergistic effect in the majority of TNBC cell lines, with synergy demonstrated in PARPi-resistant cell lines, and the strongest synergy demonstrated in a clinically-relevant model, a BRCA2-mutant cell line with acquired resistance to olaparib. Furthermore, significant dose reductions for both LNT1 and PARPi suggest the potential for an effective targeted combination with less toxicity.

Similar to what was previously reported, we demonstrated that FEN1/EXO1 inhibition demonstrated greater sensitivity in BRCA1-mutant cell lines. Since BRCA1-mutant cell lines are dependent on EXO1 function [36], we also evaluated EXO1 expression. We showed that the sensitivity of LNT1 may correlate with a high ratio of EXO1:FEN1 expression in our panel of TNBC cell lines. However, 2/3 antagonistic cell lines were also BRCA1-mutant (MDAMB436 and SUM149PT). This may be driven in part by an intrinsic upregulation of mismatch repair, DNA replication, E2F targets, and G2/M checkpoint signalling pathways.

In the BRCA2-mutant HCC1395 cell line, we observed high sensitivity to talazoparib but more resistance to LNT1. However, FEN1 was shown to be synthetically lethal in BRCA2-mutant ovarian and colon cancer cells [18], primarily due to defects in base excision repair, increased replication stress from impaired okazaki fragment processing, and compromised double-strand break repair via microhomology-mediated end-joining. While one can speculate that perhaps a higher FEN1:EXO1 ratio may be required for LNT1 to demonstrate such synthetic lethality, we did find that LNT1 was synergistic with PARPi, which is concordant with what was previously reported. Due to the high potency of talazoparib with its PARP-DNA trapping capacity, it is plausible that the PARP-single strand break intermediates require base excision repair, and thus FEN1 for resolution [18].

In the context of BRCA^WT cells, while studies have shown correlations between FEN1 and PARP-1 expression with response to PARPi or FEN1 inhibition [37,38], we did not identify such relationships. FEN1 regulation is complex, with multiple interacting proteins, post-translational modifications, and localization in various subcellular compartments [39]. Thus, total RNA or protein expression may not be the best predictive biomarker. BT549 is deficient in PTEN which causes significant replication fork stalling [40], and can explain its lower endogenous fork speed. Together with a mutation in ATR [21], BT549 was the only cell line to increase fork speed by targeting FEN1 alone. In this context, unligated Okazaki fragments can promote PARP-1 activity [38], allowing PARPi to strongly increase fork speed and DNA damage. While we did not observe an impact of the combination on cl-PARP expression, this is similar to what we previously observed with the synergistic combination of PARPi and carboplatin in BT549, since there may be other pathways through which PARPi can induce cell death [7,21]. In a similar manner, MDAMB231 is ATM-mutant, which may allow for potentiation with LNT1 and PARPi, due to impairment of DNA damage response pathways [41].

To better understand the differences between the synergistic and other cell lines, we evaluated the treatment effects on fork speed and DNA damage. In the synergistic cell lines, BT549 and HCC1395-OlaR demonstrated higher fork speed than MDAMB231 and MX1. BT549 and HCC1395-OlaR also demonstrated significant increases in percentage of γ-H2AX+ cells at 72 h with the combination of PARPi and LNT1, when there was less induction of DNA damage with the single agents. While this trend was maintained for the HCC1395-OlaR cell line at 10 days, BT549 demonstrated a dose-dependent increase in the 53BP1 product score at 10 days, suggesting the need for cells to cycle to observe DNA damage response when targeting FEN1/EXO1. However, talazoparib induced high levels of DNA damage both at 72 h and at 10 days with MDAMB231, with a similar response identified with HCC1806, suggesting that high levels of DNA damage by talazoparib may dampen the benefit achieved in combination with FEN1/EXO1 inhibition. There is rationale to explain the enhanced DNA damage in the synergistic cell lines. In response to UV irradiation, phosphorylated FEN1 migrates from the nucleolus, where it is accumulated, to the nuclear plasma to rescue stalled DNA replication forks, and translocated back to nucleolus upon DNA repair [42]. Hence, it is probable that the sustained DNA damage induced by PARPi depletes nucleolar FEN1, thereby amplifying the DNA damage response and cell death.

Finally, there may be another mechanism which can explain the synergy amongst the PARPi-resistant cell lines. We demonstrated that mismatch repair and ribosomal pathways were upregulated in PARPi-resistant cell lines, similar to what was previously reported in mouse mammary cancer cell lines and tumors [34,43]. While PARP-1 binds to small nucleolar RNAs that stimulate PARP-1 activity, PARPi impedes ribosomal DNA transcription and ribosome biogenesis [44]. FEN1 knockdown was also shown to be involved in downregulating ribosomal RNA processing and ribosome biogenesis in HEK293T cells [45]. Therefore, it is plausible that inhibition of ribosome biogenesis may play an important role in the synergy observed with talazoparib and LNT1.

We also demonstrated that FEN1 overexpression was associated with a poorer survival in breast cancer patients. These results are similar to what was previously reported in breast cancer and other cancer types [[46], [47], [48], [49]]. The synergy that we identified with FEN1/EXO1 inhibition and PARPi is in alignment with what was previously reported in breast cancer cell lines when combining a FEN1 inhibitor with radiation, another type of DNA damaging therapy [46]. This combination approach was associated with a marked inhibition of cell survival in a BRCA2-mutant cell line and BT549, with an important induction of micronuclei with BT549.

Taken together, targeting FEN1/EXO1 with PARPi is an effective combination strategy to increase replication fork speed, DNA damage, cell cycle arrest, and cell death. Importantly, this combination approach demonstrated synergy in TNBC cell lines with both intrinsic and acquired resistance to PARPi. Since current FEN1 inhibitors have been difficult to use in-vivo due to low potencies and pharmacokinetic profiles [50], our study supports further investigation in the development of new anti-FEN1 compounds. Targeting FEN1/EXO1 and PARP-1/2 demonstrates great potential to provide a new targeted combination approach for TNBC patients, leading to decreased toxicity and improved survival.

Funding information

This research project was funded by the Fonds de recherche du Québec – Santé Operating Grant for Young Clinician Investigators, File No 265384, Institut de Cancer de Montréal, and the Scotiabank Chair in the diagnosis and treatment of breast cancer. The following salary awards were attributed: S.H. by the Fonds de recherche du Québec for the Chercheur-Boursier Clinicien; Institut de Cancer de Montréal for the Canderel Bursaries to M.F., D.A, and E.F., and CRCHUM for the Postdoctoral Bursary in breast cancer to M.F.

CRediT authorship contribution statement

Mallory I. Frederick: Writing – review & editing, Writing – original draft, Visualization, Methodology, Investigation, Formal analysis. Elicia Fyle: Writing – original draft, Methodology, Investigation, Formal analysis. Anna Clouvel: Investigation, Formal analysis. Djihane Abdesselam: Investigation. Saima Hassan: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

SH receives speaker bureau fees from AstraZeneca and Merck to discuss the role of genetic testing in breast cancer patients. This has no influence on the conduct of the reported study. MF, EF, AC and DA have no financial interests to disclose.