Synergistic Cellular Toxicity from Inhibition of Poly(ADP-ribose) Glycohydrolase (PARG) and Ubiquitin-Specific Protease 1 (USP1)

Abstract

Ubiquitin-specific protease 1 (USP1) is an emerging target for poly(ADP-ribose) polymerase 1 (PARP1) inhibitor-resistant and BRCA1/BRCA2 mutant tumors. USP1 is a deubiquitylating enzyme responsible for the removal of the mono-ubiquitin mark on FANCD2, PARP1, and the replication factor proliferating cell nuclear antigen (PCNA), among other proteins. USP1 facilitates proper PCNA-mediated polymerase switching from error-prone trans-lesion synthesis DNA polymerases to replicative DNA polymerases. Due to the critical role of USP1 in DNA synthesis and DNA repair, and the discovery that USP1 deubiquitylates PARP1, USP1 inhibitors (USP1i) were found to have a synthetic lethal relationship with PARP1 inhibitors (PARPi), suggesting a mechanistic link between poly(ADP-ribose) (PAR) dynamics and USP1-mediated ubiquitin hydrolysis. However, the relationship between USP1 inhibition and inhibitors of poly(ADP-ribose) glycohydrolase (PARGi), the primary enzyme responsible for PAR hydrolysis, has not been resolved. Using cell cytotoxicity, synergy, PCNA-ubiquitin, and PAR analyses, it is demonstrated herein that PARG inhibition, combined with USP1 inhibition, leads to increased levels of mono-ubiquitinated PCNA, decreased PAR accumulation, and synergistic cytotoxicity between ML323, a potent USP1i, and PDD00017273, a model PARGi. Future studies will focus on the mechanism that contributes to USP1/PARG synthetic lethality, the mechanism of cell death, and the impact of USP1 on PAR/ubiquitin dynamics and replication stress signaling.

1. Introduction

1.1. Base Excision Repair and Poly(ADP-ribose) Signaling

Base excision repair (BER) is a DNA repair pathway that resolves damaged DNA bases to maintain genomic stability. The canonical BER pathway is responsible for repairing non-helix-distorting lesions caused by spontaneous DNA decay or exogenous mutagenic sources in all phases of the cell cycle [1]. Examples of such DNA lesions include nucleobase deamination, oxidation, or methylation [1,2]. The pathway is initiated by a lesion-specific DNA glycosylase, which recognizes modified bases [1,3], flips out the damaged nucleoside, and cleaves the glycosidic bond [1,3]. This cleavage creates an apurinic or apyrimidinic (AP) site in the DNA. After the excision of the modified base, an incision step is performed to remove the ribose sugar and, in some instances, its corresponding 5′ phosphate group [1,4,5,6]. There are three main pathways for this incision step, depending on the type of DNA lesion and repair requirements. Two of the incision pathways involve glycosylase-mediated β- or β,δ-elimination [3]. The third, and primary, BER incision step relies on apurinic/apyrimidinic endonuclease 1 (APE1) to hydrolyze the DNA backbone 5′ to the AP site, leaving a 3′-hydroxyl and a 5′-deoxyribose phosphate (5′-dRP) residue [5,7,8,9].

Following incision of the DNA backbone, poly(ADP-ribose) polymerase isoforms 1 and 2 (PARP1 and PARP2) are activated by the resulting single-stranded DNA break, leading to synthesis of poly(ADP-ribose) (PAR) polymers covalently bound to PARP1/PARP2 and to adjacent proteins, that functions as a signaling platform to recruit DNA repair proteins to the vulnerable single-stranded DNA break site [6,10]. The PAR-binding and PAR-recruited proteins include DNA repair, DNA damage signaling, metabolic, and DNA/RNA binding proteins, among others [11]. A prominent PAR-binding, and PAR-modified, protein is XRCC1 [11,12,13], a BER scaffold protein that subsequently recruits the XRCC1 binding proteins DNA polymerase β (Polβ), DNA ligase isoform 3 (LIG3), and APTX, to facilitate the completion of BER [10,14,15]. To replace a single nucleotide, short-patch BER relies on Polβ and the template strand to re-incorporate a non-damaged nucleotide [2,5,16]. Long-patch BER replaces a gap of 2–12 nucleotides around the original lesion site [1,17]. Polβ and other DNA polymerases, in concert with proliferating cell nuclear antigen (PCNA), synthesize a new strand of DNA that displaces the lesion site and creates a 5′ overhang of single-stranded DNA [1,18]. This overhang is excised by flap endonuclease 1 (FEN1), and the gap is subsequently ligated [19,20]. Little is known about the determining mechanism that differentiates between long-patch and short-patch BER [21]. Subsequently, the PAR scaffold is hydrolyzed, primarily by the PAR-degrading enzyme poly(ADP-ribose) glycohydrolase (PARG) [22], to then complete BER. Importantly, failure to degrade PAR leads to persistent binding of BER proteins to the lesion site, promoting cell cycle arrest and cell death [6,10,15,23,24].

In replicating cells, PARP1 and BER play a prominent role in Okazaki fragment processing [25]. In cancer cells undergoing replication stress, PARP1/PARP2 are activated in response to the accumulation of DNA lesions, R-loops [26], unusual DNA structures (hairpins, triplexes, G-quadruplexes) [23,27,28,29,30,31,32,33,34,35,36,37,38,39], and single-stranded DNA gaps [28,40,41,42]. The subsequent synthesis of PAR leads to the recruitment of XRCC1 and related BER proteins [23,24]. PARG, via its interaction with PCNA at the replication fork, counteracts prolonged replication stress by degrading the PAR polymer [23,37,43,44]. PARG inhibitors (PARGi) thereby cause accumulation of PAR at the replication fork [23,25], and induce the activation of the ATR/CHK1 S-phase checkpoint [23,24,28,36,37,45,46,47,48]. In turn, XRCC1 and XRCC1-dependent BER factors regulate the level of PAR at the replication fork, modulating cellular sensitivity to inhibitors of PARP1/PARP2 and PARG [23,24]. Ultimately, failure to complete BER in replicating cells promotes the accumulation of PAR and cytotoxic single-strand and double-strand DNA breaks associated with replication catastrophe [24].

1.2. Ubiquitin-Specific Protease 1 (USP1)

Ubiquitin-specific protease 1 (USP1), also known as ubiquitin carboxyl-terminal hydrolase 1 (Ubp1), is a member of the deubiquitylating enzyme family (DUBs) [49]. USP1 was originally identified as the DUB that modulates the level of monoubiquitylated FANCD2 and FANCi to regulate the Fanconi anemia pathway [50,51,52] and was found to be responsible for the removal of mono- and poly-ubiquitin tags from PCNA, regulating PCNA stability and trans-lesion DNA synthesis (TLS) [53,54]. USP1 contains the conserved catalytic triad domain: Cys90, His593, and Asp751. By itself, USP1 is minimally active; however, once it dimerizes with USP1 associating factor 1 (UAF1), the heterodimer is fully active for deubiquitylation of target proteins [55,56]. Interestingly, ATPase family AAA domain-containing protein 5 (ATAD5, the human homolog of the Saccharomyces cerevisiae protein ELG1) targets the USP1-UAF1 complex for deubiquitylation of PCNA [57].

In the context of TLS, USP1 mediates conformational regulation of PCNA for proper polymerase loading and replication processivity. During the S-phase of the cell cycle, the bidirectional replication machinery may encounter bulky DNA lesions. To prevent asymmetry of fork opening and elongation of single-stranded DNA, PCNA, a homotrimer sliding clamp, will recruit error-prone, low-fidelity DNA polymerases to swiftly overcome mutagenic lesions. Once the replication machinery stalls after encountering a lesion, RAD18 will mono-ubiquitylate PCNA at K164 to switch on low-fidelity TLS polymerases [58]. Poly-ubiquitylation will occur via UBE2K at K63 if the lesion is particularly bulky [58]. Once the lesions are bypassed, PCNA can reload high-fidelity DNA polymerases onto the DNA following USP1-mediated deubiquitylation to prevent the accumulation of mismatches during normal replication [59].

In the context of replication stress, USP1 targets HELLS [60], recently found to be involved in the cellular response to alkylating agents that damage DNA [61], and TRIP13 [62], shown to respond to mutant KRAS-mediated replication stress in pancreatic cancer cells [63]. Other USP1 targets include TRAF2 [64], SAR1A [65], and ATG14 [66], among others [59,67]. Importantly, USP1 also impacts PAR signaling by targeting PARP1 [68,69], regulating PARP1 stability [69], the efficacy of PARP-inhibitor mediated PARP1-trapping [68], and synergizing with PARP-inhibitors [70] to overcome PARP-inhibitor resistance [71].

1.3. Synthetic Lethality Cancer Treatment

In the context of cancer treatments, the BER pathway presents a range of molecular targets to exploit the intrinsic genomic instability of cancer cells for selective therapies [72,73]. Cancer cells rely on suppressed DNA repair mechanisms for tumor evolution; however, inhibition of compensatory DNA repair/DNA damage response [14] mechanisms can lead to selective killing [74]. One such target that has established success in a clinical setting is PARP1, giving rise to numerous clinically tractable PARP inhibitors (PARPi) [27,75]. Targeting PARP1 for inhibition works particularly well in homologous recombination (HR) deficient tumor cells because they target a compensatory pathway selective to these cancer cells [76,77,78]. PARPi function by inhibiting PARP1 and PARP2 enzymatic activity, trapping PARP enzymes on the DNA at lesion sites, and triggering replication fork collapse and cell death [79,80,81]. Healthy cells can tolerate effective doses of PARPi because the HR pathway is intact [82]. This concept is known as synthetic lethality [83], defined as cell death due to the inactivation of two gene pathways only in conjunction [84]. Cancer cells rely on genetic mutations to overcome barriers to unregulated proliferation; however, these mutants become over-reliant on existing, unmutated, and compensatory genome maintenance pathways for survival. Therefore, these intact pathways can be targeted via drug-mediated inhibition, causing selective killing of cancer cells. This concept has developed into a frontier of oncology [85]. Nevertheless, synthetic lethality presents many drawbacks and challenges—one of which includes resistance due to selective tumorigenic mutations, epigenetic change, or cellular/molecular adaptation [86].

Previous work has uncovered that PARG is often upregulated in many cancers [23,45]. PARG represents the complementarity of PARP1/PARP2. Whereas PARP1 and PARP2 mediate PARylation signaling via PAR biosynthesis, PARG facilitates the hydrolysis and removal of the PAR polymers [87]. The delicate balance of this enzyme-mediated PAR polymerization and depolymerization at DNA lesion sites is critical for timely DNA repair. Therefore, due to its connection to PARP and upregulated status in cancer cells, PARG inhibitors (PARGi) have emerged as therapeutic options to PARPi-resistant tumors in clinical trials. PDD00017273, used in this study, is one of many PARGi currently available [88].

Although PARGi are just emerging in pre-clinical studies and in clinical trials [23,24,28,89,90,91], there is an unmet need to develop viable therapeutic treatments to target prospective PARGi-resistant tumor cells. We have shown that PARGi-mediated cellular toxicity is dependent on cellular NAD⁺ levels [23,91], on defects in BER [23,24], on activity of the checkpoint proteins CHK1 or ATR [24,92], or the arginine methyltransferases PRMT1 or PRMT5 [24]. Previous studies report that USP1 is upregulated and associated with poor prognosis in many cancer types, including osteosarcoma, colorectal, non-small cell lung, breast, and gastric cancers [67,93,94]. Therefore, due to its implicated role in cancer resistance and reported super-additive or synergistic relationship in combination with PARPi, we investigated the relationship and functional interactions between inhibitors of PARG and USP1.

In this study, ML323 was used as a potent USP1 inhibitor (USP1i) [95]. Isolation treatment studies suggest USP1 inhibition in BRCA1 mutant cells results in decreased nuclear PCNA levels and consequently decreased DNA synthesis and increased basal levels of S-phase-specific DNA damage [58,96]. Previous reports also indicate that USP1 binds to PARP1 as a substrate and removes K63-linked ubiquitin tags [68]. USP1 is suggested to regulate PARP1 chromatin trapping and PARylation activity due to this USP1/PARP1 interaction [68]. This molecular interaction helps explain the established synergistic relationship between PARPi and USP1i and success in using USP1i to overcome PARPi resistance in HR-deficient cell lines [68,70].

Nevertheless, there has been little to no investigation on the molecular relationships of USP1i/PARGi combination therapies. Therefore, this study aimed to characterize the synergistic effects and underlying molecular mechanisms of USP1i/PARGi combination treatment in cancer cell lines. We hypothesize that simultaneous inhibition of USP1 and PARG will exert a synergistic cytotoxic effect in cancer cells, accompanied by increased levels of ubiquitinated PCNA and disrupted PARylation signaling.

2. Materials and Methods

2.1. Chemicals and Reagents

DMEM and RPMI-1640 were from Corning (Manassas, VA, USA). MEM, L-glutamine, hygromycin, and penicillin/streptomycin were from Gibco (Grand Island, NY, USA). PDD00017273 (PARG inhibitor; Sigma-Aldrich, St. Louis, MO, USA) and ML323 (USP1 inhibitor; Selleckchem, Houston, TX, USA) were dissolved in DMSO to prepare a stock solution at a concentration of 100 mM and stored at −30 °C.

Table 1. Reagent List.

	Source	Identifier
Antibodies
Rabbit anti-Ubiquityl-PCNA (Lys164) (Immunoblot—1:1000)	Cell Signaling Technology (Danvers, MA, USA)	Cat #13439
Rabbit anti-PCNA (Immunoblot—1:1000)	Novus Biologicals (Centennial, CO, USA)	Cat #SY12-07
Rabbit anti-α-Actinin (Immunoblot—1:1000)	Cell Signaling Technology	Cat #3134
Goat anti-rabbit HRP-conjugate (Immunoblot—1:3000)	Bio-Rad (Hercules, CA, USA)	Cat #1662408EDU
Chemicals
Heat-inactivated fetal bovine serum	Bio-Techne (Minneapolis, MN, USA)	Cat# S11150H
Penicillin/streptomycin	Thermo Fisher Scientific (Waltham, WA, USA)	Cat# 15140-122
DMEM	Corning (Corning, NY, USA)	Cat# 15-017-CV
Trypsin-EDTA	Thermo Fisher Scientific	Cat# 25200-056
Blotting grade non-fat dry milk	Bio-Rad	Cat# 170-6404
NuPage 4–12% Bis-Tris gel	Invitrogen (Carlsbad, CA, USA)	Cat# NP0323BOX
Clarity Western ECL Substrate	Bio-Rad	Cat# 1705060
SuperSignal West Femto Maximum Sensitivity Substrate	Thermo Fisher Scientific	Cat# 34095
RPM1 1640	Corning	Cat# 31724011
Formaldehyde solution (4%)	Thermo Fisher Scientific	Cat# 242845
DPBS	Corning	Cat# 25324005
ML323	Selleckchem	Cat #S7529
MNNG	Sigma-Aldrich	Cat# 129941
NRH	NuChem (Québec, Canada)	Custom
PDD00017273	Sigma-Aldrich	Cat# SML1781
Vectashield	Vector Laboratories (Newark, CA, USA)	Cat# H-1000-10
Trypan Blue Stain 0.4%	Invitrogen	Cat# 2216894
Hoechst 33342	Thermo Fisher Scientific	Cat# 6224910
Propidium iodide	Sigma (St. Louis, MO, USA)	Cat# P4170
Cell Lines
U2OS	ATCC (Manassas, VA, USA)	Cat# HTB-96
ES-2	ATCC	Cat# CRL-1978
LN428	Dr. Ian Pollack (University of Pittsburgh)	[91,97,98]
Software and Algorithms
ImageJ	Image J 1.48v	https://imagej.net/ij/download.html (10 December 2025)
Adobe Illustrator	Adobe Systems (San Jose, CA, USA)	Version 29.3
NIS-Elements	Nikon Instruments (Tokyo, Japan)	Versions 4.51 and 5.11
GraphPad Prism	GraphPad (La Jolla, CA, USA)	Version 9, (Mac OS X)
SynergyFinder 3.0	[99]	10.1016/j.gpb.2022.01.004
BioRender (Toronto, ON, Canada)	Graphical Abstract	https://app.biorender.com

provides a comprehensive list of chemicals and reagents used in this project.

2.2. Cells and Cell Culture

ES-2 and U2OS cells were obtained from ATCC. ES-2/XRCC1-KO cells were developed by us previously [24]. LN428 cells were a kind gift from Dr. Ian Pollack (University of Pittsburgh) and have been described by us previously [91,98]. U2OS/LivePAR cells were described by us previously [48,100]. ES-2 cells were cultured in McCoy’s 5A medium supplemented with 10% heat-inactivated fetal bovine serum and penicillin/streptomycin. LN428 cells were cultured in MEM Alpha medium supplemented with 10% heat-inactivated fetal bovine serum, L-glutamine, and penicillin/streptomycin/amphotericin. U2OS cells were cultured in DMEM supplemented with 10% heat-inactivated fetal bovine serum, penicillin/streptomycin, and L-glutamine. All parental and modified cell lines were grown in tissue culture incubators at 37 °C and 5% CO₂.

2.3. Cell Viability Analysis

Cell viability in response to drug treatments was assessed according to the following protocol [101], as we have described previously [48]. Cells were initially seeded into each well of a 96-well plate at a density of 800 cells per well. After a 24 h incubation period, cells were exposed to either a single or combined dosage (with multiple dilutions as specified in the figures), without the removal of preconditioned media. Subsequently, after 120 h (5 days), the total cell population was identified by staining with Hoechst 33342 (2 µM), and dead cells were identified by staining with propidium iodide (1.5 µM), followed by a 15 min incubation at 37 °C. Enumeration of both total and dead cells was performed using the Celigo S Image Cytometer (Nexcelom Bioscience, Perkin Elmer, Shelton, CT, USA). This involved capturing the Hoechst dye signal (excitation/emission wavelength for the blue channel, 377 nm/470 nm) and the propidium iodide signal (excitation/emission wavelength for the red channel, 531 nm/629 nm).

2.4. Immunoblots

Whole-cell lysates (40–50 μg protein) were applied to precast NuPAGE^® Novex^® 4–12% Bis-Tris gels (Invitrogen; Carlsbad, CA, USA) and electrophoresed for 1 h at 100–120 V (1.5 h), as described [23]. Following electrophoresis, the separated proteins were transferred to a nitrocellulose membrane using a Turboblotter (Bio-Rad). The membrane was initially blocked with B-TBST (TBS buffer containing 0.05% Tween-20 and supplemented with 5% blotting-grade non-fat dry milk; Bio-Rad) for 1 h at room temperature. Subsequently, the membrane was incubated overnight at 4 °C with primary antibodies in B-TBST. Table 1 provides details on the primary antibodies and their respective dilutions. After washing, the membranes were treated with secondary antibodies in B-TBST for 1 h at room temperature. The secondary antibodies used were Bio-Rad anti-rabbit-HRP conjugate. Following additional washing steps, the membrane was exposed to a chemiluminescent substrate, and protein bands were visualized using a Bio-Rad Chemi-Doc MP imaging system (Hercules, CA, USA). To determine the induction factors (I.F.) of mono-ubiquitylated PCNA) a densitometry analysis of the immunoblots was performed using ImageJ. First, the relative amounts of a-actinin were determined to account for differences in loading. Then, taking into account these differences, the relative amounts of mono-ubiquitylated PCNA in the treated samples were expressed as a factor compared to the control.

2.5. LivePAR Assay (Fluorescence-Based Detection of Poly(ADP-ribose) Polymers)

This assay was designed to detect poly(ADP-ribose) (PAR) polymers at DNA damage sites, as described by us previously [48,100]. U2OS/LivePAR cells (cells modified to express the LivePAR probe) were cultured with DMEM media containing 45,000 mg/L of glucose, L-glutamine, sodium bicarbonate, FBS, and penicillin–streptomycin.

Experimental procedure: Coverslips cleaned with diethyl ether and then stored in ethanol were placed in 60 mm Petri dishes. U2OS/LivePAR cells (250,000) were then seeded in 4 mL of DMEM media covering the coverslips. One Petri dish was used for each treatment condition. After incubation (24 h), the cells were either left untreated or treated with 20 µM ML323, 100 µM NRH + 10 µM MNNG + 10 µM PDD0017273, or 20 µM ML323 + 100 µM NRH + 10 µM MNNG + 10 µM PDD0001723. The cells were first pre-exposed to ML323 for 30 min before the other reagents were added. The cells were then incubated for 60 min with the other reagents. Following the incubation, the media was removed and washed with Dulbecco’s Phosphate-Buffered Saline (DPBS). The cells were then fixed twice, first with formaldehyde, followed by a 7:3 methanol: acetone solution [48,100]. The coverslips with the fixed cells were then transferred onto a glass microscopy slide containing Vectashield. The fluorescence signals for EGFP (LivePAR) and DAPI (DNA) were detected, imaged, and quantified using a Nikon Ti2-E inverted confocal microscope at a 63 X (N. A. 1.42) magnification. The microscope is equipped with Ax-R and Ax-R 2k Resonant + Galvo Scan Head. Data analysis: The fluorescent PAR foci were counted, in addition to the number of nuclei, using an in-house script and Image J, as described [48,100]. The number of PAR foci per nucleus in the captured image was reported as a function of the specified treatment conditions. A Mann–Whitney U test was performed to determine the statistical difference between PAR foci counts.

2.6. Statistics

Data are shown as the mean ± standard deviation from 2 to 3 independent experiments. Student’s t-test was used for comparisons between the two groups. Two-way ANOVA was used for multiple comparisons. The significance between the control and experimental groups is indicated by p-values. p-values are indicated by asterisks with * p < 0.05. GraphPad PRISM V10.5.0 was used for statistical analysis.

3. Results

3.1. Synergistic Interplay Between PARG Inhibition and USP1 Inhibition

The first objective of this study was to determine whether USP1i and PARGi have a synergistic relationship. To accomplish this task, cell viability assays were performed with the USP1i ML323 and the PARGi PDD00017273. Cell treatments were conducted in isolation (dose response) and then in combination. Synergistic calculations using standard mathematical models were then employed to determine the level of synergism across two separate cell lines.

In addition to testing the response in the ES-2 ovarian cancer cell line, we also tested the response in the isogenic BER-deficient ES-2/XRCC1-KO cells [24], used here as a model for BER deficiency due to XRCC1’s critical role in recruiting BER proteins to damage sites [14,24,102]. Notably, the killing curves (Figure 1) demonstrate different cellular responses to PARGi and USP1i. The ES-2/XRCC1-KO cells are more sensitive to PARGi treatment (Figure 1A), in line with our recent reports on BER-deficiency giving rise to increased response to PARGi [23,24,48]. Interestingly, we note a small but measurable difference in the response to USP1i treatment when comparing the WT and XRCC1-KO cells (Figure 1B). This may provide greater insight into the mechanisms of cell death from USP1 inhibition in future studies. Since XRCC1 is implicated in the resolution of single-stranded DNA breaks, greater sensitivity in the KO cells indicates that PARGi, in the absence of XRCC1, causes incomplete single-strand break resolution and, therefore, the intolerable accumulation of lethal double-stranded breaks in replicating cells [23,24]. Since the killing curve for the USP1i treatment (Figure 1B) displays a slight difference in cell death +/− XRCC1, this may also suggest that XRCC1 loss may render cells slightly dependent on USP1 function, at least in part. However, the small increase in sensitivity of the ES-2/XRCC1-KO cells to USP1i treatment may also suggest that PARGi and USP1i affect different cellular pathways or induce cell killing via separate mechanisms.

3.2. Mathematical Models for Drug Synergy Analysis

Figure 1C provides a visual representation of the enhanced killing effect of the PARGi/USP1i combination treatment, also demonstrated by the IC₅₀ of each experiment with PARGi and USP1i in combination (Figure 1D). Synergy calculations were then performed using the ZIP, Loewe, Bliss, and HAS models to quantify the enhanced combinatory toxicity. The Zero Interaction Potency Model (ZIP) first assumes that the two drugs investigated do not interact [103]. If this were the case, then the potency of each drug should remain the same even in combination. The ZIP begins by modeling the single-drug treatment using the following formula:

$$y_1={\displaystyle \frac{({\frac{x}{m_1})}^{{\lambda }_1}}{1 + {\left(\frac{x}{m_1}\right)}^{{\lambda }_1}}}$$

(1)

The m value represents the midpoint of the curve—the dose represented by x, which outputs 50% of the maximum value of y. The $\lambda$ parameter represents the potency or Hill parameter of the curve, giving rise to the steepness of the curve. Assuming drug independence, the drug combination treatment should only raise the baseline of the y₁ curve without altering the potency parameter. Therefore, this new combination equation can be represented as the following, where y₂ follows the same form as y₁:

$${ y}_{1\leftarrow 2}={\displaystyle \frac{y_2 + ({\frac{x}{m_1})}^{{\lambda }_1}}{1 + {\left(\frac{x}{m_1}\right)}^{{\lambda }_1}}}$$

(2)

However, if there is an observed combination effect, the potency parameters should be altered. This new observed combination effect can be modeled as the following, where m_1→2 and λ_1→2 are the projected potency and shape parameters for drug 1 when adding x₂:

$$y_{c 1\leftarrow 2}={\displaystyle \frac{y_2 + ({\frac{x_1}{m_{1\leftarrow 2}})}^{{\lambda }_{1\leftarrow 2}}}{1 + {\left(\frac{x_1}{m_{1\leftarrow 2}}\right)}^{{\lambda }_{1\leftarrow 2}}}}$$

(3)

The delta score can then be calculated by taking the deviation between $y_{1\leftarrow 2}$ and $y_{c 1\leftarrow 2}$, where $y_{c 2\leftarrow 1} \mathrm{a}\mathrm{n}\mathrm{d} y_{2\leftarrow 1}$ represent the dose response curves for drug 1 in combination with drug 2.

$$\delta \left(\theta \right)={\displaystyle \frac{y_{c 1\leftarrow 2}-y_{1\leftarrow 2}}{2}}+{\displaystyle \frac{y_{c 2\leftarrow 1}-y_{2\leftarrow 1}}{2}}$$

(4)

The greater the $\delta$ value, the greater the deviation between the adjusted combination curves and the zero-interaction curves [103].

The Loewe model assumes that a drug cannot interact with itself [104]. Therefore, if two drugs have the same mechanism-of-action, then the combined dose should simply match the single-dose response instead of having a higher baseline as assumed by ZIP. The Loewe model uses the following equation to determine whether there is synergy between two drugs:

$${\displaystyle \frac{x_1}{X_{LOEWE}^1}} + {\displaystyle \frac{x_2}{X_{LOEWE}^2}}=1$$

(5)

In a Cartesian coordinate system with the y- and x-axes defined as the concentration of drug 1 and 2, respectively, Equation (5) represents an isobole of additivity that is a straight line connecting the intercepts $X_{LOEWE}^1$ and $X_{LOEWE}^2$ together. The terms $X_{LOEWE}^1$ and $X_{LOEWE}^2$ represent the doses of drugs 1 and 2 in isolation that produce the following equation:

$$y_{LOEWE}={\displaystyle \frac{E_{min} + E_{max}{\left(\frac{x_1 + x_2}{m}\right)}^{\lambda }}{1 + {\left(\frac{x_1 + x_2}{m}\right)}^{\lambda }}}$$

(6)

E_min and E_max represent the minimum and maximum effects of the drug. If the combination curve y_c is greater than y_LOEWE, then X_c¹ > X_LOEWE¹ and X_c² > X_LOEWE². This changes Equation (5) to take the following form:

$${\displaystyle \frac{x_1}{X_c^1}}+{\displaystyle \frac{x_2}{X_c^2}}<1$$

(7)

The differences between the isobole of additivity Equations (5) and (7) can then be analyzed to determine the degree of synergy [104].

The Bliss model assumes a stochastic process between the two drugs [104]. The expected combination effect is calculated based on the probability of independent events.

$$g_{Bliss}=g_1\left(x_1\right)+g_2\left(x_2\right)-g_1\left(x_1\right)g_2\left(x_2\right)$$

(8)

The equation of the general form g₁(x_i) can be defined as the conditional response curve of the form 1 − f_i(x_i), where f_i(x_i) is the original single treatment dose–response. If the drugs work independently, the probability to survive can be represented as

$$\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{b}\mathrm{a}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y} \mathrm{o}\mathrm{f} \mathrm{S}\mathrm{u}\mathrm{r}\mathrm{v}\mathrm{i}\mathrm{v}\mathrm{a}\mathrm{l}=\left(1-g_1\left(x_1\right)\right)\left(1-g_2\left(x_2\right)\right)$$

(9)

The expected combined inhibition is then

$$g_{Bliss}=1-\left(1-g_1\left(x_1\right)\right)\left(1-g_2\left(x_2\right)\right)=g_1\left(x_1\right)+g_2\left(x_2\right)-g_1\left(x_1\right)g_2\left(x_2\right)$$

(10)

To calculate the synergy score, the deviation between the observed inhibition of the combination treatment and the calculated g_Bliss is taken [104].

Finally, the Highest Single Agent or HSA model assumes that the expected combination effect for non-synergistic drugs is the maximum effect observed by either drug alone (the drug that possesses the greater maximum effect). Any additional maximal effect observed within the combination treatment is treated as synergy [103].

Since the synergy values represented in Figure 2A (and shown in Figure 2B left) are all above the standard threshold of 10 for the four mathematical models (ZIP, Loewe, HAS, Bliss), these cell viability experiments support the synergistic relationship between ML323 and PDD0017273. The sensitivity scores (Figure 2B right) represent another quantitative method to compare synergy findings across cell lines and experiments. The relative inhibition (RI) values are listed (Figure 2B) for the single dose treatments, representing the overall inhibitory effect of a drug. The RI values are calculated by taking the area under the dose–response curve as a percentage of the maximum possible area (100% inhibition). Therefore, an RI of 15.91 for ML323 indicates that ML323 achieves 15.91% of the maximum possible inhibition over the given dose range. The Combination Sensitivity Score (CSS) measures how effective the combination drug treatment is at inhibiting cell growth or viability. The score reports, on average, the extent to which the combination treatment reaches its full inhibitory potential [105]. A CSS of 69.15 indicates that the combination treatment achieves 69.15% of the maximum possible inhibitory effect.

Figure 3 represents a confirmation analysis of the increased cell killing when combining the USP1i and the PARGi, shown here in the LN428 glioma cell line. As shown, the LN428 cells are more sensitive to the cell-killing effect of the PARGi (Figure 3A) than the ES-2 cell line (Figure 1A). Interestingly, the LN428 cell line demonstrated slight resistance to the USP1i (Figure 3B) as compared to the ES-2 cell line (Figure 1B). The treatment conditions were essentially the same, with adjustments to selected treatment dose concentrations. This experiment, in the LN428 cell line, was performed to increase the validity of the synergy results, showing an increased response to the USP1i when exposed to minimally toxic doses of the PARGi (Figure 3B,C). The synergy scores, above 10 for all four models (Figure 4), confirm the synergistic conclusion from the ES-2 cell experiments (Figure 2). The synergy calculations point to the combination’s heightened toxicity to these cancer cells. Although the synergy is diminished in the LN428 experiments, this could be due to unknown compensatory mechanisms within the LN428 cell line (Figure 4) that may tolerate the drug combination more so than the ES-2 cell line (Figure 2). Regardless, the data confirm the hypothesis that USP1i and PARGi have a synergistic relationship with regard to cytotoxicity.

3.3. USP1 and PARG Inhibition Alters PCNA Ubiquitination Status

The cell viability assays confirmed the hypothesized synergistic relationship between PARGi and USP1i. With this finding, the second objective of the study was to begin to understand the molecular mechanism underlying this synergy. To accomplish this, immunoblot assays were employed to test for mono-ubiquitylated PCNA levels in the USP1i-treated, PARGi-treated, and combination-treated cells. USP1 inhibition leads to elevated cellular levels of ubiquitylated PCNA [58]. Therefore, measuring levels of ubiquitin-tagged PCNA can provide insight into the efficacy of USP1i treatment. By comparing mono-ubiquitylated PCNA levels in cells treated with the inhibitors in isolation and combination, we aimed to determine whether the combination treatment would exacerbate dysregulation of PCNA modifications and, consequently, replication stress. This molecular link could potentially explain the synergistic killing effect observed in the cell viability assays.

In both the LN428 and ES-2 cell lines (Figure 5), the relative levels of mono-ubiquitylated PCNA increase when treated with the USP1 inhibitor ML323 alone and there is a further increase in mono-ubiquitinated PCNA when cells are treated with ML323 in combination with the PARG inhibitor PDD00017273. One interpretation of these findings suggests that the inhibition of PARG, which is known to cause S-phase arrest, leads to greater induction of replication stress including polymers of ADP-ribose, PAR, at single-stranded DNA break sites and at sites of replication fork stalling [23,24]. These markers of DNA replication stress and DNA damage may then require the facilitation of TLS during S-phase to swiftly overcome these lesion sites. An increased level of ubiquitylated PCNA may be required to deal with the PARGi-induced replication stress; however, failure to properly remove this ubiquitin signal, following exposure to ML323, leads to the potential degradation of PCNA and total replication fork collapse.

3.4. Poly(ADP-ribose) Polymers Decrease with USP1i Treatment

As a member of a deubiquitylating class of enzymes, UPS1 not only targets PCNA but also PARP1 [68,69]. In these earlier reports, it was demonstrated that USP1 inhibition impacts PARP1 stability and the cellular response to PARP1 inhibitors [68,69]. Therefore, we next tested whether the USP1i/PARGi combination treatment impacted the PARylation pathway.

To evaluate the capacity for PARP1/PARP2-induced PARylation in cells, we used the LivePAR assay to assess PAR formation activity mediated at damaged DNA sites induced by the alkylating agent methyl-nitronitroso-guanidine (MNNG) [15,100,106,107]. Here, we used U2OS cells transduced with a lentivirus to express the LivePAR probe, enhanced green fluorescence protein (EGFP) that is fused to the C-terminus of the WWE domain of RNF146 (amino acids 100–182), a protein domain that binds to poly(ADP-ribose) chains [15,100,106,107] (Figure 6). In response to DNA damage, PARP1 is activated and synthesizes poly(ADP-ribose) (PAR) at sites of genomic DNA breaks [100]. As we have shown previously, cells expressing this EGFP-RNF146 (100–182) fusion accumulate at sites of PARylation, forming EGFP-foci, and any defects in PARylation can be quantified [48].

The inhibition of USP1, in addition to PARG, showcases the importance of USP1 in the recruitment of repair proteins to sites of DNA damage. The decreased level of PAR foci in the combination treatment (Figure 6) supports the synergistic cell killing observed by the combination treatment. These results suggest that USP1/PARG dual inhibition leads to the disruption of multiple DNA repair and replication stress mechanisms, effectively preventing the activation of compensatory survival pathways that ultimately promote the observed enhanced cytotoxicity.

The negative control experiments show that negligible PAR foci are formed when PARG is active, since PARG hydrolyzes these polymers. However, inhibition of PARG combined with DNA-damaging agents (MNNG) hyperactivates PARP1/PARP2, as indicated by the increased level of fluorescent foci (Figure 6). The fluorescent foci represent PAR chains at DNA strand breaks, serving as a signal for DNA repair protein recruitment [1,15]. The PAR chains are visible since these cells express a PAR-binding protein fused to the fluorescent protein EGFP [15,100,106,107]. Here, we found that U2OS/LivePAR cells treated with the USP1 inhibitor ML323 + PARG inhibitor resulted in decreased PAR foci levels compared to the positive control treatment with PARG inhibitor PDD00017273 alone. These results align with previous literature reporting USP1’s role in regulating PARP1 deubiquitylation and stability [68,69]. The ubiquitin labelling of PARP1 serves as a critical regulator of PARP1 activity, and, therefore, PAR chain formation and USP1 inhibition leads to the accumulation of polyubiquitin chains on PARP1’s CAT domain [68]. This covalent tagging creates an extended conformation that sterically hinders conformational shifting necessary for PARP1 to conduct PARylation once bound to single-stranded DNA breaks. Further, USP1 inhibition impacts the stability of PARP1 [69], suggested to be targeted for ubiquitylation by Iduna (RNF146) [108].

4. Discussion

This study set out to explore the molecular interplay between USP1i and PARGi and the resulting cytotoxic response. This was executed to elucidate the toxicity of each inhibited pathway and to help eventually introduce this combination treatment to a clinically relevant setting. The initial cell viability experiments using ES-2 and LN428 cell lines confirmed the hypothesized synergistic relationship between ML323 and PDD00017273. This hypothesis was based on previous literature that reported synergy between USP1i and PARPi [58]. The synergy calculations, using four mathematical models in two cell lines, all met the threshold of defined synergy.

Besides determining synergy, the second objective of this study was to understand molecular mechanisms underlying the enhanced cytotoxic effects of the dual treatment, with a focus on the ubiquitylation of PCNA, and alterations in PARylation. While it may be possible that the increase in cell death observed is via an elevation in the level of apoptotic signaling, the cells in this study are p53 defective and so it is feasible that other mechanisms of cell death, such as autophagy, necrosis, necroptosis, among others [109], are also possible and so will be considered in future studies. These studies were done not only to provide a rational experimental approach towards investigating synergy in a general sense but also to enhance understanding of cytotoxic effects of drugs implicated in BER inhibition or disruption. This could potentially provide insight into future drug design for cancer treatments. Immunoblot experiments were performed in the same two cell lines to gauge the level of mono-ubiquitylated PCNA. The results showed an increase in this post-translational modification of PCNA in the combination treatment. This result provided the initial insight explaining the synergy. PARG inhibition facilitates the increase in PAR chains at single-stranded DNA breaks created not only at damaged DNA sites but also during DNA replication surrounding Okazaki fragment processing and the response to replication-stress induced PARP1/PARP2 activation [23,24,25,28]. The increased encounters of these PAR chains with the polymerase machinery likely facilitate greater reliance of TLS to maintain polymerase processivity, thereby explaining the increased levels of mono-ubiquitylated PCNA to avoid replication fork collapse. However, the failure of PCNA to return high-fidelity polymerases, due to USP1 inhibition, results in increased replication stress and potentially proteasome-mediated PCNA degradation [58,96].

Importantly, an interaction between USP1 and PARP1 was previously reported, implicating USP1 in the regulation of PARP1 enzymatic activity and cellular PAR dynamics [68,69]. Therefore, it was important to investigate how the combination treatment affected PAR chain formation, further complicating the proposed mechanistic insight concluded from the immunoblot assays. The results of the LivePAR assay showed that the USP1i/PARGi combination treatment decreased PAR foci in the U2OS/LivePAR cell line as compared to PARGi treatment alone. This was likely due to decreased activity of PARP1 from USP1 inhibition. Therefore, in one interpretation of the results, the combination treatment leads to an overall increase in replication stress from two sources. The first could be from PARGi-mediated replication fork stalling via an increase in PAR chains over normal levels. The second source could be related to the slight yet significant decrease in PAR chains, which signal the suppression of existing DNA repair pathways that fail to respond to the replication stress caused by PARG inhibition. Overall, these results paint a larger molecular picture explaining the synergy in response to the combined inhibition of USP1 and PARG.

5. Conclusions

Overall, our results suggest that USP1i/PARGi cytotoxic synergy is most likely originating from the heightened induction of replication stress due to the loss of both USP1 and PARG function. Therefore, investigating how USP1 and PARG regulate replication stress in these cell lines would be useful to further elucidate this mechanism, including biomarkers such as phosphorylation of CHK1, RPA, or KAP1, as we have recently reported [24,48], as well as DNA fiber assays to visualize replication alterations following USP1i or PARGi treatments, as we have shown in response to PARGi treatment [23]. It is possible that the two inhibitors synergize in response to the accumulation of replication-stress associated single-strand DNA gaps and activation of the S-phase checkpoint, in line with recent reports on inhibitors of both USP1 [110,111] and PARG [23,24,28]. Similarly, the strong sensitivity of cells deficient in the BER factor XRCC1 (XRCC1-KO) to the PARGi PDD00017273 and the slight sensitivity of XRCC1-KO cells to the USP1i ML323 also supports the likelihood that the synergy is related to replication stress markers such as the accumulation of single-strand DNA gaps regulated by BER [24]. Moving forward, mass spectroscopy proteomic analysis, to understand the full ubiquitylation status of PCNA and PARP1, as well as new USP1 targets, may provide a more complete picture regarding the role of USP1 and PARG in replication stress signaling and PAR/ubiquitin dynamics.