PARP1 expression predicts PARP inhibitor sensitivity and correlates with metastatic potential and overall survival in melanoma

Metastatic melanoma is still a difficult‐to‐treat cancer type owing to its frequent resistance mechanisms to targeted and immunotherapy. Therefore, we aimed to unravel novel therapeutic strategies for melanoma patients. Preclinical and clinical studies show that melanoma patients may benefit from a treatment with poly (ADP‐ribose) polymerase (PARP) inhibitors (PARPi). In this study, we focus on PARP1 as a potential biomarker to predict the response of melanoma cells to PARPi therapy. We found that melanoma cells with high basal PARP1 expression exhibit significantly increased cell death after PARPi treatment owing to higher PARP1 trapping compared with melanoma cells with low PARP1 expression. In addition, we could demonstrate that PARP1 expression levels are low in nonmalignant skin cells, and metastatic melanomas show considerably higher PARP1 levels compared with primary melanomas. Most strikingly, we found that high PARP1 levels correlate with worse overall survival of late stage metastasized melanoma patients. In conclusion, we show that PARP1 might act as a biomarker to predict the response to PARPi therapy, and that in particular the late stage metastasized melanoma patients are especially sensitive to PARPi therapy owing to elevated PARP1 expression. Our data suggest that the PARPi cytotoxicity primarily will affect the high PARP1 expressing melanoma cells, rather than the low PARP1 expressing nonmalignant skin cells resulting in only low side effects.

PARP inhibitors, with PARP inhibitors emerging as a potentially beneficial therapy for patients with late-stage metastasized melanoma in particular.

| INTRODUCTION
Malignant melanoma, a skin cancer arising from mutated melanocytes, is a very aggressive tumor and patients have poor prognosis because of a high potential for metastasis to distant organs. 1 At the moment, patients with mutated mitogen-activated protein kinase (MAPK) signaling pathway profit from a combination therapy of BRAF inhibitors and MEK inhibitors. 2 Despite the good efficacy of the targeted therapy, several molecular mechanisms that lead to acquired resistance after long-term treatment restrict the positive outcome. 3Melanoma patients also benefit from checkpoint inhibitors; however, the success of immunotherapy is limited owing to tolerability and response rate issues. 4Therefore, it is of major importance to investigate possible new therapeutic options for melanoma patients.
Poly (ADP-ribose) polymerase (PARP) inhibitors (PARPi) are the first FDA and EMA approved anti-cancer drugs to directly target the DNA damage response.PARP1 is a major regulator of the DNA repair mechanism termed base excision repair (BER).By acting as nicotinamide adenine dinucleotide (NAD + ) analog, PARPi block the PARylation process of PARP1, which results in inefficient BER.Moreover, PARP1 can no longer auto-PARylate itself, resulting in stalled replication forks by the inability of PARP1 to dissociate from the DNA and ultimately the induction of double strand breaks (DSB), which exert the major cytotoxic function of PARPi therapy.This phenomenon is termed as PARP trapping effect. 5cently we showed that PARPi effectively reduce the growth of MAPK inhibitor resistant melanoma cells and synergize with MAPK inhibitors through a synthetic lethal interaction highlighting the importance of PARPi treatment for melanoma patients in a clinical setting. 6rthermore, patients with a homologous recombination repair (HRR) deficiency (HRD), and thereby the inefficiency to repair DSB, particularly profit from PARPi therapy. 7However, additional biomarkers might even further predict the response to PARPi therapy, independent of HRD.Therefore, we aimed to unravel the influence of PARP1 expression levels in melanoma cells on PARPi therapy response.Furthermore, we wanted to find out whether PARP1 expression levels might have an impact on melanoma progression and prognosis.
All experiments were performed with mycoplasma-free cells. 8The cell lines A375 and SKMel28 show protein expression of the most common HRR genes including Rad51, BRCA1, BRCA2, and XRCC2. 9In addition, the cell lines A375, SKMel28, 451Lu, WM1552, WM3211, and WM793 did not show damaging mutations in ATM, ATR, BRCA1, BRCA2, CHEK2, FANCA, FANCB, FANCC, FANCD2, FANCE, FANCF, FANCC, FANCL, and PALB2 (information from Depmap portal) and are therefore considered HRR proficient.Melanocyte isolation from human foreskin was approved by the ethics committee of the medical faculty of the University Tübingen (654/2014BO2) and performed according to the principles of the Declaration of Helsinki.After removal of surplus fatty and vascular tissue, the foreskin was cut into small 1 cm 2 pieces and incubated overnight at 4 C in epidermal keratinocyte medium with supplements (CELLnTEC) with 10 μg/mL gentamicin and 0.25 μg/mL amphotericin B (CELLnTEC) containing 10 mg/mL Dispase II (Roche) to digest the basal lamina.The next day, epidermis and dermis were carefully separated, and small slices of the epidermis were incubated in 0.05% trypsin-EDTA (Merck Millipore) for 30 min, while small slices of the dermis were incubated in 1 mg/mL collagenase A (Roche) in CnT-40 Medium (CELLnTEC).Digestion was stopped using RPMI1640 medium (Thermo Fisher Scientific) containing 10% FBS (Biochrom), and single cells were obtained using a 100-μm-pore-size cell strainer (Corning Incorporated).After centrifugation, cells were resuspended in CnT-40 medium (CELLnTEC).

| Viability analysis and inhibitors
Cell viability was analyzed with the 4-methylumbelliferyl heptanoate (MUH) assay as previously described. 6Cells were treated with talazoparib (Medchemexpress, HY16106) or olaparib (Medchemexpress, HY-10162) for 72 h.Experiments were performed in at least three biological replicates and technical quintuplicates.

| RNA isolation and RT-qPCR
RNA isolation was performed with the help of the Nucleospin ® RNA Kit (Macherey-Nagel 740955) according to the manufacturer's protocol.The cDNA was produced and RT-qPCR was performed as previously described. 6RT-qPCR was performed with the Lightcycler ® 96 Instrument (Roche).The following primer sequences were used: PARP1 forward: CCAAGCCAGTTCAGGACCTCAT, PARP1 reverse: GGATCTGCCTTTTGCTCAGCTTC. Actin forward: CACCATTGG-CAATGAGCGGTTC, Actin reverse: AGGTCTTTGCGGATGTCCACGT.
Actin served as a housekeeping gene.Experiments were performed in technical triplicates and at least three biological replicates.

| Transfection
The melanoma cells A375 and SKMel28 were transiently transfected with either siRNA against PARP1 or with a PARP1 expressing plasmid (Origene, RC207085).The following siRNA against PARP1 (siPARP1, Dharmacon, D-006656-02-0005) was used.For the transfection of the siRNA against PARP1, Lipofectamine RNAimax (Thermofisher Scientific™, 13778075) and for the PARP1 expressing plasmid, Lipofectamine3000 (Thermofisher Scientific™, L3000150) was used according to the manufacturer's protocol and as previously described. 624 h after the transfection process, cells were either harvested for RNA and protein expression analysis or reseeded into 96 well plates for further MUH analysis.Experiments were performed in at least three biological replicates.

| Migration and invasion assay
Boyden chamber-based migration and matrigel invasion assays were performed as previously described. 6Twenty-four hours after transfection of SKMel28 and A375 with either PARP1 overexpression plasmid or siRNA against PARP1, 1 Â 10 5 and 2 Â 10 5 cells, respectively, were seeded onto the transwell inserts and after 4 h migrated or invaded cells were fixed and stained.Experiments were performed in biological triplicates and nine images per group were quantified.

| Database analysis
The Kaplan-Meier curve of the TCGA SKCM 470 database of melanoma patients was generated with R2 genomics analysis and visualization platform (https://r2.amc.nl).The median was used to divide the patients into high PARP1 expression and low PARP1 expression and a survival curve for each group with a cut-off at 10 years was plotted.The GSE112509 dataset was used to compare the PARP1 gene expression between melanocytic nevi and primary melanomas and the GSE8401 as well as the TCGA SKCM 375 dataset was used to compare the PARP1 gene expression between primary and metastatic melanomas.For correlation analysis of the respective datasets the DepMap portal was used (https://depmap.org/portal/).
For analysis of the metastatic potential and the penetrance, the MetMap 500 dataset was utilized.Penetrance was defined as the percentage of mice detected with the cell line after cancer cell injection. 10For the PARP1 gene expression, the DepMap Public 23Q2 dataset was used and for the talazoparib and olaparib drug sensitivity, the PRISM Repurposing 19Q4 dataset was utilized.

| Statistical analysis
All experiments were statistically analyzed using GraphPad Prism version 9.1.2.Data that are statistically significant (p < .05)were labeled with asterisks (* for p < .05,** for p < .01,*** for p < .001,and **** for p < .0001).Unless otherwise stated, statistical analysis was performed by unpaired t-test when two groups were compared with each other.
For the analysis of the correlation between to datasets, a simple linear regression model was used.For the analysis of survival curves the R2 platform was used to compare the high PARP1 expression and low PARP1 expression groups.The median was used to divide high PARP1 expression, and low PARP1 expression groups and a log rank test was performed to check for significant differences in the overall survival of the two groups.

| PARP1 expression levels are predictive for overall survival and PARPi response in melanoma
To analyze the potential role of PARP1 as a biomarker in melanoma, we correlated overall survival of melanoma patients with PARP1 expression levels in their tumors.Interestingly, overall survival analysis showed a significant difference in the prognosis of melanoma patients depending on PARP1 expression levels.Melanoma patients with high PARP1 expression have a significant lower overall survival probability compared with patients with low PARP1 expression (Figure 1A).We then split these patients into early stage patients without metastasis (stage 0-II) and late-stage patients suffering from metastasis (stage III-IV).Interestingly, while no significant difference in overall survival between high and low PARP1 expressing patients in stage 0-II melanoma patients was seen, metastasized patients in stages III-IV with high PARP1 expression levels show significant worse outcome than late-stage patients with low PARP1 expression levels (Figure 1B,C).To unravel the influence of PARP1 expression on PARPi response, we performed cell viability assays after PARPi treatment (olaparib or talazoparib) of melanoma cells, which express PARP1 at different levels.Indeed, a significant correlation between PARP1 mRNA or protein expression and PARPi sensitivity could be detected, with cells having high PARP1 expression showing a significantly better response to PARPi therapy (Figure 1D and Figure S1A).
Consistent with this, analysis of a correlation between the PRISM Repurposing 19Q4 and DepMap Public 23Q2 dataset checking for the drug sensitivity (olaparib and talazoparib) and PARP1 gene expression of various melanoma cells in vitro validated our results (Figure 1E).RT-qPCR analysis as well as immunoblot analysis revealed that non-malignant human melanocytes show lower PARP1 gene expression levels compared with metastatic human melanoma cells, which is significant on protein level (Figure 1F,G).In addition, we found a gradual increase in PARP1 gene and protein expression between melanoma cells in radial growth phase (RGP), vertical growth phase (VGP) and metastatic melanoma cells (Figure 1F,G).
In line with this, we could demonstrate that melanocytic nevi show significant reduced PARP1 gene expression levels compared with primary melanoma cells (Figure 1H).
To conclude, we found that in melanoma patients high PARP1 expression correlates with worse overall survival, especially in the metastasized patients and that the sensitivity toward PARPi positively correlates with PARP1 gene expression in melanoma patients and in melanoma cell lines.These data implicate, that especially late stage metastasized patients with high PARP1 expression levels would profit from PARPi therapy with limited side effects on low PARP1 expressing benign cells.

| PARP1 trapping after PARPi treatment induces melanoma cell death
As the main cytotoxicity of PARPi therapy is known to be dependent on PARP trapping and thereby the induction of DSB, we were interested, if PARP1 expression levels correlate with trapped PARP1.

Indeed, cells with high PARP1 gene expression and thereby increased
PARPi sensitivity, such as A375 (see Figure 1D and Figure S1B), show significantly enhanced chromatin bound PARP after talazoparib treatment, whereas cells with low PARP1 gene expression and thereby reduced PARPi sensitivity, such as SKMel28 (see Figure 1D and Figure S1B), show a lower and nonsignificant increase in PARP1 trapping after PARPi treatment (Figure 2A).To support the hypothesis, that PARPi therapy induces higher PARP trapping and thereby more DSB in high PARP1 expressing cells, we performed siRNA induced knockdown of PARP1 in A375 cells and subsequently treated them with PARPi.Strikingly, in high PARP1 expressing A375 cells, the knockdown of PARP1 significantly desensitized these cells to olaparib and talazoparib treatment (Figure 2B and Figure S1C).Vice versa, an upregulation of PARP1 gene expression via plasmid transfection led to a sensitization of low PARP1 expressing SKMel28 cells to olaparib and talazoparib treatment (Figure 2C and Figure S1D).
These data indicate that PARP1 expression correlate with a higher PARP trapping effect after PARPi therapy and a higher killing efficiency.

| PARP1 expression level correlates with metastatic potential of melanoma cells
To analyze the influence of PARP1 expression on the metastatic potential of melanoma cells, we analyzed PARP1 gene expression in primary versus metastatic melanoma samples in two datasets.Interestingly, we found a significant lower RNA expression of PARP1 in the primary melanoma group compared with the metastatic melanoma group (Figure 3A).When analyzing the role of PARP1 in metastasis formation and metastatic penetrance of melanoma cells, a trend was observed, confirming that PARP1 might positively influence the metastatic potential of melanomas (Figure 3B,C).Furthermore, we checked for differences in the migratory and invasive potential of high PARP1 expressing A375 and low PARP1 expressing SKMel28 cells.The migratory and invasive potential of high PARP1 expressing A375 melanoma cells was enhanced compared with low PARP1 expressing SKMel28 melanoma cells (Figure 3D,E and Figure S1C,D).However, neither upregulation of PARP1 in low PARP1 expressing SKMel28 cells, nor downregulation of PARP1 in high PARP1 expressing A375 cells significantly changed their migratory or invasive capacity (Figure 3D,E).Therefore, our data suggest that PARP1 is more likely involved in the metastatic process of melanoma cells.

| DISCUSSION
This study shows the importance of basal PARP1 expression on sensitivity toward PARPi therapy.Elevated PARP1 levels were associated with higher PARP1 trapping after PARPi treatment leading to high cytotoxicity in these cells.Upregulation of PARP1 sensitized low PARP1 expressing cells to PARPi therapy, and vice versa, downregulation of PARP1 desensitized high PARP1 expressing melanoma cells to PARPi treatment.We found that melanoma cells show significantly increased PARP1 gene expression levels compared with non-malignant skin cells.Moreover, our data revealed that high PARP1 expression levels positively correlate with the metastatic potential of melanoma cells.Most strikingly, we were able to show that high PARP1 levels are associated with worse overall survival exclusively in late-stage melanoma patients.
Currently, PARPi are FDA approved for the treatment of HRD cancers. 11Our data indicate that elevated PARP1 levels result in increased PARP trapping by PARP inhibitors, and thereby higher cytotoxicity.Consistent with another study, we here show that PARP1 therefore could serve as a useful biomarker to forecast PARPi response in cancer patients. 12,13We thus propose that, in addition to testing the HRD status of patients before receiving PARPi therapy, PARP1 expression levels should be analyzed to further predict PARPi therapy success.
In this study we demonstrate, that PARP1 expression levels correlate with increased metastatic potential of melanoma cells.5][16][17] Interestingly, together with other publications, we additionally found that PARP1 expression is significantly increased in melanomas compared with nonmalignant skin cells, such as melanocytes or melanocytic nevi. 17With this, we suggest that undesirable side effects of PARPi therapy targeting noncancerous cells should be comparably low.Our data indicate, that PARP1 is associated with a worse overall survival of melanoma patients.8][19][20][21] To our knowledge, this is the first publication to show that the decreased survival probability of high PARP1 expressing cancers is dependent on the tumor stage, and that only in late-stage metastatic melanoma patients (stage III-IV), a significant difference in overall survival between low and high PARP1 expressing patients is evident.
Taken together, our study reveals that PARP1 might be an effective biomarker to forecast PARPi response.We suggest that patients with late-stage metastatic melanoma should be screened for PARP1 expression.Patients who exhibit high PARP1 expression should additionally be screened for HRD, as we expect a poor overall survival.We suggest that patients with high PARP1 expression and an HRD should receive PARPi therapy.Together with our previous results, we propose that patients displaying high PARP1 levels but are proficient in HRR should profit from a combination of PARPi therapy plus MAPKi therapy, as MAPKi therapy in BRAF mutated melanoma cell lines induces an HRD phenotype and thereby act synthetic lethal in combination (see graphical abstract). 6

AUTHOR CONTRIBUTIONS
Birgit Schittek contributed to conceptualization, funding acquisition, investigation, methodology, supervision, writing-original draft, writing- F I G U R E 3 PARP1 expression levels correlate with metastatic potential of melanoma cells.(A) GSE8401 dataset (left) and TCGA SKCM 375 (right) was used to compare PARP1 gene expression between 23 primary melanomas and 57 metastatic melanomas (left) or 65 primary melanomas and 266 metastatic melanomas (right).Unpaired t-test was used to compare the two groups.(B,C) The relative metastatic potential in brain, lung, liver, bone, and kidney using the MetMap 500 data compared with PARP1 expression (B) and the penetrance in brain, lung, liver, bone, and kidney (C) using the DepMap Public 23Q2 of melanoma cells is shown.The correlation between the relative metastatic potential and the PARP1 expression was analyzed using the simple linear regression model.(D,E) Migration and invasion assay using SKMel28 PARP1 overexpressed (PARP1OvE) and control (Ctr.) as well as A375 PARP1 knockdown (siPARP1) and control (Ctr.) cells is shown.The mean number of invaded (D) or migrated (E) cells per image section is shown.Nine images per group were analyzed.Experiments were performed in biological triplicates and one-way ANOVA with multiple comparison was used to compare the groups.

F
I G U R E 1 PARP1 levels are predictive for overall survival and poly (ADP-ribose) polymerase (PARP) inhibitors (PARPi) response in melanoma.(A-C) Kaplan Meier curve of overall survival of 459 melanoma patients (A) of the TCGA SKCM 470 dataset divided into stage 0, I, Ia, Ib, IIa, IIb, IIc (B) and stage III, IIIa, IIIb, IIIC, and IV (C) grouped into high and low PARP1 expression.Median was used to define the cutoff between high and low PARP1 expression.R2 was used to plot the graph and perform statistical analysis.(D) 4-Methylumbelliferyl heptanoate (MUH) cell viability assay was performed after treatment of melanoma cells with different concentrations of olaparib or talazoparib for 72 h.IC50 was calculated.PARP1 gene expression normalized to ß-actin of the respective cells is shown.The correlation between the IC50 levels and the PARP1 expression was analyzed using the simple linear regression model.(E) PARPi drug sensitivity using the PRISM Repurposing 19Q4 data compared with PARP1 expression using the DepMap Public 23Q2 is shown.The correlation between talazoparib or olaparib drug sensitivity and the PARP1 expression was analyzed using the simple linear regression model.(F,G) PARP1 gene expression (F) or PARP1 protein expression (G) of PARP1 in six melanocytes, two melanoma cell lines in radial growth phase (RGP), two melanoma cell lines in vertical growth phase (VGP), and six metastatic melanoma cell lines is shown.All experiments were analyzed in four biological replicates.One-way ANOVA with multiple comparison was used to compare the groups.(H) GSE112509 dataset was used to compare PARP1 gene expression between 23 melanocytic nevi and 57 primary melanomas.Unpaired t-test was used to compare the two groups.

F
I G U R E 2 PARP1 trapping after Poly (ADP-ribose) polymerase (PARP) inhibitors (PARPi) treatment induces melanoma cell death.(A) Immunoblot analysis of chromatin bound fraction of A375 and SKMel28 cells after the treatment with 0.005% methyl methane sulfonate (MMS) plus 2 μM talazoparib (+) or untreated cells (À) for 6 h.An exemplary picture of the immunoblot is shown (left side).Relative PARP1 protein expression (ratio PARP1 and H2AX) of three biological replicates is shown in the graph (right side).(B,C) 4-Methylumbelliferyl heptanoate (MUH) cell viability assay of SKMel28 and A375 melanoma cell lines treated with different concentrations of PARPi (olaparib and talazoparib) for 72 h.PARP1 knockdown (siPARP1) and control (Ctr.) cells (B) or PARP1 overexpression (PARP1OvE) and control (Ctr.)(C) transfected cells were used.The treatment started 24 h after the transfection process.Comparison of fits of the nonlinear regression was performed to analyze statistical differences between the untransfected and transfected cells.Relative viability at 25 μM PARPi treatment (ratio of Ctr. vs. siPARP1/ PARP1OvE) of 4 biological replicates (B) and three biological replicates (C) is shown.
review and editing.Lisa Marie Fröhlich contributed to conceptualization, wet lab experiments, formal analysis, investigation, methodology, visualization, writing-original draft, writing-review and editing.Ana Villar Miyar performed wet lab experiments, formal analysis, investigation, methodology, and visualization.Tamara Heintze performed wet lab experiments, formal analysis, investigation, methodology, visualization, and writing-review and editing.Birgit Sauer performed wet lab experiments, methodology, and visualization.The work reported in the article has been performed by the authors, unless clearly specified in the text.