PARP Inhibitor Activity Correlates with SLFN11 Expression and Demonstrates Synergy with Temozolomide in Small Cell Lung Cancer Free

Lung cancer is the most common cause of cancer death in both men and women, responsible for more deaths than colon cancer, breast cancer, and prostate cancer combined. Small cell lung cancer (SCLC) represents approximately 15% of all lung cancers. SCLC is strongly associated with tobacco exposure and is expected to be an increasingly prevalent concern over the next several decades, particularly in Asia and the developing world, mirroring smoking trends. Worldwide, more than 200,000 people die from SCLC every year. Overall 5-year survival for patients with SCLC is a dismal 7%, with almost all survivors being among the minority diagnosed with limited stage disease.

All standard therapies for SCLC are DNA damaging agents, inducing either covalent DNA adducts and crosslinks [cisplatin, carboplatin, temozolomide (TMZ)], or single-strand or double-strand breaks (ionizing radiation, etoposide, topotecan, irinotecan; refs. 1–3). Despite its grim survival statistics, SCLC is also notable for being remarkably responsive to combinations of these DNA damaging agents. In fact, about 25% of patients with limited stage disease are cured with concomitant cisplatin, etoposide, and radiation. The exceptional sensitivity to DNA-damaging therapies may relate to the underlying genetics driving SCLC oncogenesis. Nearly 100% of cases of SCLC have homozygous loss or inactivation of both RB transcriptional corepressor 1 (RB1), encoding the primary regulator of the G₁–S cell-cycle checkpoint, and tumor protein p53 (TP53), critical for multiple DNA damage response pathways (4). The notable sensitivity of SCLC to DNA damage suggests that targeted inhibition of the DNA repair pathways operant in SCLC is a particularly attractive strategy and can substantially augment the efficacy of therapies in our current armamentarium.

Poly-(ADP)-ribose polymerase enzymes (PARP) function to detect and mark DNA single-strand breaks (SSB) by binding to the site of DNA damage and synthesizing poly-(ADP)-ribose chains, which recruit a host of scaffold proteins and DNA repair enzymes to resolve the break (5). poly(ADP-ribose) polymerase 1 (PARP1) protein levels are upregulated in SCLC relative to other lung cancers, and initial studies suggested that this upregulation was associated with increased sensitivity of SCLC lines to PARP inhibitors in vitro (6). These promising preclinical data in SCLC supported the inclusion of a SCLC cohort in a phase I study of the PARP inhibitor talazoparib with promising initial signals of efficacy (7).

PARP1 was first described as a regulator of base excision repair, but has since been implicated in the function of homologous recombination (HR), non-homologous end joining, and microhomology-mediated end joining (5, 8). There are at least two distinct mechanisms of action of PARP inhibitors: enzymatic inhibition and the more recently recognized PARP trapping (9, 10). Inhibition of PARP enzymatic activity was initially thought to explain the synthetic lethality observed with PARP inhibitors in breast and ovarian cancers with BRCA1, DNA repair associated (BRCA1) and BRCA2, DNA repair associated (BRCA2) mutations. These mutations lead to deficiencies in HR, leaving these cancers highly dependent on PARP-mediated repair (11, 12). The PARP inhibitor olaparib was recently approved for treatment of germline BRCA1/2 mutant ovarian cancer patients based on the results of a randomized phase II study (13). In addition, the FDA granted olaparib Breakthrough Therapy designation for treatment of patients with castrate-resistant metastatic prostate cancer harboring BRCA1/2 and ATM serine/threonine kinase (ATM) mutations. However, mutations in BRCA1/2 are notably rare (≤ 3%) in SCLC based on recent comprehensive genomic analyses (4, 14).

PARP trapping is a distinct mechanism of action of PARP inhibitors, whereby the inhibitor/PARP complex becomes fixed on the DNA at sites of SSBs, leading to a failure to repair, and, with replication, induction of multiple double strand breaks. PARP trapping may be responsible for synergy between PARP inhibitors and DNA damaging agents that increase the prevalence of SSBs. Further, this mechanism may be operant in cancers without defined HR deficiencies (9). The various PARP inhibitors in clinical development and clinical use vary in relative potency for both enzymatic inhibition and PARP trapping effects. Olaparib and talazoparib have comparable levels of catalytic inhibition, while talazoparib is ∼100-fold more potent than olaparib at trapping PARP–DNA complexes (9, 10). Rucaparib appears to have activity similar to olaparib, while veliparib is less potent both in enzymatic inhibition and in trapping activity (9, 10).

Beyond inactivating mutations in known mediators of HR such as BRCA1/2, other mechanisms may result in HR deficiency in sporadic tumors, including epigenetic silencing of BRCA1 and HR pathway disruptions in other known and unknown mediators of this pathway (15, 16). This has led to substantial interest in strategies for defining “BRCAness,” or HR deficiency (HRD), including using characteristic patterns of mutation and loss from whole-exome sequencing data to generate HRD scores (17–19). Discovering novel mechanisms of HRD in sporadic tumors may broaden the therapeutic potential of PARP inhibitors.

In this study, we sought to define determinants of PARP inhibitor activity in SCLC, and also to evaluate the combinatorial activity of PARP inhibition with the DNA damaging agent TMZ. Recent studies have suggested that TMZ is a particularly effective agent in recurrent metastatic SCLC, with both systemic and central nervous system activity, leading to its inclusion in the National Comprehensive Cancer Network guidelines for standard treatment of this disease (20). We report schlafen family member 11 (SLFN11) as a critical determinant of PARP inhibitor sensitivity in SCLC. SLFN11 was recently reported to be actively recruited to sites of DNA damage, and to inhibit HR, strongly supporting these findings (21). We found the association between PARP trapping and cytotoxic activity is stronger in SCLC than in other tumor types, and focused subsequent analyses on the strongest PARP trapper, talazoparib. By clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR associated protein 9 (Cas9) gene editing and shRNA approaches, we established that loss of SLFN11 confers resistance to talazoparib in SCLC cell lines. In vivo, we confirmed that SLFN11 expression by immunohistochemistry (IHC) is associated with tumor response to talazoparib in multiple patient-derived xenograft (PDX) models. We also demonstrated marked synergy between talazoparib and TMZ in multiple SCLC cell lines, and that combinatorial treatment with talazopoarib and TMZ extends the spectrum of activity beyond those tumors with high SLFN11 levels. Collectively, our results demonstrate that SLFN11 is a relevant predictive biomarker of sensitivity to PARP inhibitor monotherapy in SCLC and support combinatorial therapy with TMZ as a promising therapeutic strategy in SCLC.

SCLC cell lines were purchased from American Type Culture Collection and maintained as recommended. All cell lines tested negative for mycoplasma and were short tandem repeat verified/authenticated within 6 months of use. Dimethyl sulfoxide (DMSO) was used as a vehicle for all in vitro experiments. Talazoparib was obtained from Biomarin or Selleck Chemicals. For in vivo use, talazoparib was prepared in dimethylacetamide (DMAc; 270555, Sigma) in 1 mg/mL concentration then diluted 1:10 in 5 parts Kolliphor HS/85 parts PBS and stored at 4°C and discarded after 14 days. For in vivo dosing, talazoparib was diluted in sterile PBS to 0.2 mg/mL for weight-based dosing by oral gavage. Talazoparib vehicle was 2% DMAc/1% Kolliphor HS/97% PBS. Temozolomide (S1237; Selleck Chemicals) for in vivo use was prepared in 0.5% carboxymethylcelluse sodium salt in water for weight-based dosing by oral gavage.

Cells were collected and suspended by exposure to Accutase in 300 mL of media, then plated (5 × 10³–20 × 10³ in 42 μL) in 384-well plates (CulturPlates) 24 hours prior to treatment. Each compound was tested at 9 different concentrations (10 μmol/L to 1.5 nmol/L; DMSO concentration 0.25%), then the plates were incubated for 96 hours. The incubation was terminated by adding ATP Lite (Perkin Elmer Inc.) and luminescence was determined.

Cells were plated 24 hours before treatments in 96-well plates at 1 × 10³ to 5 × 10³ cells per well. Viability (AlamarBlue; ThermoFisher) assays were quantified on a compatible plate reader. Cell lines were treated with the indicated drugs for 5 to 7 days prior to assaying.

Primary antibodies to SLFN11 (Santa Cruz Biotechnology; sc-374339) and PARP1 (Santa Cruz Biotechnology; sc-7150) were used at a 1:250 and 1:1,000 dilution, respectively, according to the manufacturer's instructions. Primary antibodies to O-6-methylguanine-DNA methyltransferase (MGMT) (Cell Signaling Technology; #2739) and actin (Cell Signaling Technology; #4967) were used at a 1:1,000 dilution.

For quantitative infrared Western blots, all products and reagents unless specified were purchased from LI-COR Biosciences and used according to the manufacturer's recommendations (see Supplementary Methods). For film developed Western blots, enhanced chemiluminescence (ECL) detection anti-mouse or anti-rabbit HRP-conjugated secondary antibodies (GE Healthcare) were used at a 1:5,000 dilution as previously described (22). Target protein quantification was normalized relative to the highest loading control sample.

The IHC detection of SLFN11 antibody (Sigma-Aldrich; HPA023030) was performed at the Molecular Cytology Core Facility of the Memorial Sloan Kettering Cancer Center using Discovery XT processor (Ventana Medical Systems). SLFN11 primary incubation time is 5 hours, followed by 60-minute incubation with biotinylated goat anti-rabbit IgG (Vector, cat # BA-1000, 1:200 dilution), Blocker D, Streptavidin-HRP and DAB detection kit (Ventana Medical Systems) were used according to the manufacturer instructions.

Custom shRNA miR-E sequences designed to target SLFN11 were as follows:

SLFN11#1: TGCTGTTGACAGTGAGCGCTAGAAGTAATCCTTCATTTAATAGTGAAGCCACAGATGTATTAAATGAAGGATTACTTCTAATGCCTACTGCCTCGGA;

SLFN11#2: TGCTGTTGACAGTGAGCGATCAGACCAATATCCAAGAGAATAGTGAAGCCACAGATGTATTCTCTTGGATATTGGTCTGAGTGCCTACTGCCTCGGA.

A control shRNA targeting Renilla Luciferase has been previous described (23). These constructs were generated and cloned into the LT3GEPIR (pRRL) vector, which was a gift from Christof Fellmann and Johannes Zuber, then transfected into the cell lines of interest.

The lentiCRISPR v2 plasmid was a gift from Feng Zhang, Massachusetts Institute of Technology (Addgene plasmid #52961). The design of sgRNAs targeting SLFN11 was performed using publicly available software. Guide pairs were synthesized by Sigma Aldrich, annealed and inserted into the lentiCRISPR v2 backbone as described previously (24). sgRNA sequences and plasmids were confirmed by Sanger sequencing and are as follows: SLFN11#1, 1F: AGGTATTTCCTGAAGCCGAA, 1R: TTCGGCTTCAGGAAATACCTC; SLFN11#2, 2F: GAGTCCTGAGAGCAGCGCAG, 2R: CTGCGCTGCTCTCAGGACTCC.

See Supplementary Methods.

Affymetrix SNP6 array data from the CCLE were processed together, quantile-normalized, and median-polished with Affymetrix power tools (25). The birdseed algorithm was used for genotyping. PennCNV was used to generate log R ratio and B-allele frequencies (26). ASCAT was used to generate allele-specific copy number information, which was then used to compute LST, N_tAI, HRD-LOH scores in the R statistical environment (v3.0.2) following methods outlined in the initial publications (17–19).

Robust Multi-array Average (RMA) and quantile-normalized gene expression microarray data for 1,037 cancer cell lines were downloaded from the Cancer Cell Line Encyclopedia (25).

Publicly available gene expression datasets from The Cancer Genome Atlas (TCGA) were combined with our previously published genomic dataset of primary human SCLC tumors (14). The results published here are in part based upon data generated by the TCGA Research Network.

The PDXs were isolated and passaged as previously described (27, 28). All treatment and toxicity experiments were performed in female NSG mice (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ; The Jackson Laboratory) that were 6 to 8 weeks old at time of PDX injection/implantation.

Tumor volumes were calculated from manual caliper measurements with an ellipsoid formula in which volume (mm³) = (xy²)/2. Once tumor volumes reached approximately 150 mm³, mice were randomized to treatment arms and treated via oral gavage with vehicles, talazoparib (0.2 mg/kg or 0.3 mg/kg) daily, TMZ (6 mg/kg) every 4 days, or in combination (talazoparib 0.2mg/kg daily and TMZ 6 mg/kg every 4 days).

Gene expression analysis was performed using the R statistical programming environment and the Bioconductor suite of tools. Expression values for 23 SCLC samples were centered and 12,631 genes with mean absolute deviation >.3 were considered. Genes with expression correlating with mean PARP inhibitor response in vitro were identified using LIMMA to fit a linear model to each gene and generate moderated t-statistics using an empirical Bayes approach, including as covariates PARP1 gene expression, and LST score. P values were adjusted for multiple testing by the method of Benjamini and Hochberg. Spearman rank correlation performed on drug response data and HRD scores to talazoparib response.

All animal data are reported as average tumor volumes ± SD. Tumor volume comparisons between treatment groups used data from study endpoint using a two-sided Student t test; throughout the text unless indicated: *, P < 0.05 and **, P < 0.01. All graphs in figures were created using GraphPad Prism 6.0c (GraphPad Software, Inc.) or R (version 3.2.2).

Linear mixed-effect models were fit on individual animal data to assess whether response to talazoparib depends on SLFN11 expression, which was represented by the interaction between talazoparib treatment and SLFN11 expression. We used various measurements to represent SLFN11 expression in the model including IHC staining H-score, modified H-score, dichotomized H-score, Western blot, and mRNA level. The outcome was the endpoint tumor volume. The start tumor volume was included in the regression model as a covariate. Random intercepts were used to account for the clustering of animals within PDX models.

Drug synergy was determined quantitatively using the combination index (CI) method of Chou and Talalay (29). Viability was calculated across a wide range of doses for talazoparib and temozolomide individually and in combination as a constant ratio of talazoparib to temozolomide (1:10,000 and 1:2,500). CI was calculated as a function of response or fraction affected (Fa) using the formula CI = (D₁)/(D_x)₁ + (D₂)/(D_x)₂, where D₁ and D₂ are the doses used to achieve a specific response in combination and (D_x)₁ and (D_x)₂ are individual drug doses needed to achieve similar response with CompuSyn Software ver. 2005. A CI >1 indicates antagonism, while a CI <1 indicates synergism.

All animal experiments were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee of the Memorial Sloan Kettering Cancer Center.

We initially sought to determine whether previously defined assays for HRD could predict SCLC sensitivity to PARP inhibitors. The activity of cisplatin, like that of PARP inhibitors, has been previously reported to correlate with HRD scores in patients with breast and ovarian cancers (17–19). For analysis in SCLC, we evaluated the relative activity of olaparib, racuparib, and veliparib across 414 cell lines available from the Genomics of Drug Sensitivity in Cancer (GDSC) panel, including 23 SCLC cell lines (30). The observed responses to all three agents were correlated across all histologies, including SCLC (Fig. 1A).

Large chromosomal structural alterations characteristic of BRCA1/2 mutant cancers have been deployed as an indirect assessment of HRD, representing a history of “scarring” of the genome reflective of loss of this particular repair pathway. Three metrics of these HRD genomic scars are loss of heterozygosity (HRD-LOH), large-scale state transition (LST), and telomeric allelic imbalance (N_tAI; refs. 17–19, 31). These HRD measures correlate with platinum sensitivity in sporadic triple-negative breast and ovarian cancers (17–19, 32, 33). We hypothesized that HRD scores might be a relevant response discriminator to PARP inhibitors in SCLC. To interrogate this premise, we computed the three different HRD metrics (HRD-LOH, LST, N_tAI) in SCLC cell lines using publicly available Cancer Cell Line Encyclopedia (CCLE) Affymetrix SNP 6.0 array datasets (Supplementary Fig. S1 and Supplementary Table S1) (17–19, 25). Across this dataset, all three measures correlated with one another, but surprisingly, none of these three measures of HRD correlated with response to PARP inhibitor sensitivity in vitro in SCLC (Supplementary Fig. S1B). Consistent with previous work (6), we found that PARP1 is overexpressed in SCLC by gene transcript levels (data not shown). We then quantified PARP1 protein levels by near-infrared Western blotting (Supplementary Fig. S2A). Unsurprisingly, as PARP1 protein levels are almost universally high in SCLC cell lines relative to NSCLC (6), expression of PARP1 itself was not informative for response to PARP inhibition in linear regression analysis (Supplementary Fig. S2B).

We sought to identify other features that might better predict SCLC sensitivity to PARP inhibition. Using Robust Multi-array Average (RMA) data from the CCLE, we analyzed expression of 12,631 genes in 414 cell lines to identify candidate determinants of PARP inhibitor activity with potential relevance in SCLC. SLFN11 was among the top genes most significantly correlated with PARP inhibitor sensitivity, controlling for HRD score and expression of PARP1 and PARP2 (Fig. 1B; Supplementary Table S2). We observed that SCLC cell lines with high levels of SLFN11 transcript were more sensitive to PARP inhibitors and conventional cytotoxic therapy (Fig. 1C). Interestingly, across the three PARP inhibitors studied, a lower differential in activity between SCLC and other cell lines was evident in drugs with more potent PARP trapping ability (i.e., olaparib > rucaparib > veliparib; Fig. 1D).

Due to this correlation of PARP trapping potency and efficacy in SCLC, we conducted additional analyses with talazoparib, currently the most potent PARP trapper available in clinical development. We tested a panel of SCLC cell lines and found, as previously reported, stereospecific drug activity (Fig. 2A; ref. 10). We confirmed that SLFN11 transcript and protein expression correlate, by linear regression (P = 0.01; Supplementary Fig. S2C). As expected, we established that SLFN11 expression is a determinant of talazoparib sensitivity where a correlative trend by linear regression between SLFN11 protein expression and talazoparib sensitivity was observed, even in this small sample (P = 0.06; Fig. 2B).

These correlative analyses suggested that SLFN11 might be a direct determinant of PARP inhibitor activity in SCLC. Alternatively, SLFN11 expression could correlate with PARP inhibitor sensitivity, reflecting a shared association with another determining factor. To test these alternative hypotheses, we directly assessed whether SLFN11 inactivation conferred PARP inhibitor resistance by deriving SCLC cell lines with doxycycline inducible expression of 2 different shRNA sequences targeting SLFN11. We observed that knockdown resulted in a ≥ log-fold increase in the talazoparib IC₅₀, supporting a direct role for SLFN11 in drug sensitivity (Fig. 3A; Supplementary Fig. S3). To further confirm this result, we used CRISPR/Cas9 gene editing to derive SLFN11 knockout cells. We observed the same talazoparib resistance phenotype in the most effective guide RNA tested, when using a pooled selection approach to generate knockout cell lines (SLFN11-KO#2; Fig. 3B). The parental cell line and a different sgRNA sequence with inefficient knockout (SLFN11-KO#1) served as negative controls (Fig. 3B).

We have previously shown that SCLC PDX models more closely reflect their tumors of origin than standard cell line xenografts, which demonstrate gene expression alterations attributable to ex vivo cell line derivation (28). Novel therapeutic testing in PDXs in several instances appears to more accurately reflect the subsequent experience in humans (27, 34–37). These observations have recently led the NCI to announce plans to retire the NCI-60 cell line panel and replace this with a PDX repository as the preferred platform for drug discovery and testing. Therefore, we conducted in vivo testing in our available PDX models.

We characterized seven SCLC PDXs for SLFN11 protein expression by IHC staining (Fig. 4A) and conducted talazoparib efficacy experiments in each tumor model. We found talazoparib dosed daily at 0.3 mg/kg was well tolerated (Supplementary Fig. S4A). Efficacy data in vivo were entirely consistent with the cell line observations that SLFN11 expression was required for sensitivity to talazoparib. All three PDX models that were SLFN11-high demonstrated statistically significant and clinically meaningful tumor growth inhibition (TGI) with single agent talazoparib, whereas none of the four SLFN11-low models responded (Fig. 4B and C). A thoracic pathologist blinded to PDX identity and response data generated an SLFN11 H-score, which ranges from 0 to 300 and integrates three intensities of IHC nuclear staining and their frequency (Fig. 4A). We used a linear mixed-effect model to confirm the strong dependence of talazoparib efficacy on SLFN11 expression (P < 0.0001 for the interaction). The difference in the endpoint tumor volume between talazoparib and control groups increases by 6.83 mm³ for every unit increase in the SLFN11 H-score (95% confidence interval: 4.25 to 9.41 mm³; Supplementary Fig. S4C and Supplementary Table S3). Interestingly, SLFN11 H-score proved to be a stronger predictor of talazoparib efficacy across these PDX lines than either SLFN11 gene expression, or protein expression by Western blot (statistical model #1 vs. #4 and #5 in Supplementary Table S3 and Supplementary Fig. S4B and S4C). We also found a modified H-score consisting of the grouped nuclear staining intensities 2+ and 3+ categories as a robust predictor of talazoparib response (Fig. 4A; Supplementary Fig. S4D and Supplementary Table S3). These observations have practical ramifications: the ability to quantitatively assess SLFN11 by IHC staining will facilitate rapid clinical translation, allowing evaluation of formalin-fixed paraffin-embedded samples.

We evaluated gene expression levels for SLFN11 in treatment-naïve SCLC patient tumors available from our previously published work (14) and retrieved datasets from other tumor histologies available as part of TCGA. We found that SLFN11 was highly expressed in SCLC as compared with other histologies (Fig. 4D). In addition, we observed a bimodal distribution of SLFN11 gene expression (Fig. 4D, right inset, blue dashed line), that may provide meaningful stratification as a predictive biomarker in future clinical studies of SCLC.

Clinical evaluations of PARP inhibitor efficacy are increasingly shifting to combination studies focusing, in particular, on coadministration with DNA damaging agents. Of particular interest, previous studies have demonstrated combinatorial efficacy between PARP inhibitors and TMZ across multiple disease types, including colon cancer, breast cancer, melanoma, leukemia, and glioblastoma (38–42). As noted above, TMZ is of particular relevance in SCLC, as it possesses substantial activity in recurrent metastatic SCLC, including patients with brain metastases.

TMZ is a DNA alkylator that preferentially methylates the O6 position of guanine. MGMT, a demethylating DNA repair protein, reverses the O6 methylation caused by TMZ, and silencing of the MGMT gene by promoter hypermethylation has been correlated with improved tumor response and overall survival to TMZ in combination with radiotherapy in glioblastoma (43). Initial clinical data in SCLC also revealed a trend (P = 0.08) toward higher response to TMZ in patients with MGMT promoter silencing (20).

We therefore sought to evaluate the combinatorial activity of TMZ and talazoparib across our panel of SCLC models, and to assess the predictive roles of both MGMT expression and SLFN11 expression in this context. We first performed a drug synergy matrix of 35 different combinations of talazoparib and TMZ dose levels (Fig. 5A, left). We found a range of combinations between 1:70 and 1:700,000 of talazoparib to TMZ that exhibited excess response over highest single agent (HSA) activity and determined a 1:10,000 ratio of talazoparib:TMZ to be the approximate median drug ratio (Fig. 5A, right).

We then performed formal assessments of combinatorial synergy of TMZ and talazoparib using four cell lines, two with high MGMT expression (NCI-H82 and DMS114) and two with low MGMT expression (NCI-H446 and NCI-H146; Fig. 5B). We performed combination index (CI) experiments as described by Chou and Talalay (29) to quantify any synergistic interaction between talazoparib and TMZ. Cells were treated with talazoparib or TMZ alone, or a fixed ratio of 1:10,000 of talazoparib:TMZ (range tested 0.25 nmol/L:25 μmol/L to 16 nmol/L:160 μmol/L).

All four cell lines exhibited strong synergism, although in distinct dose ranges. Low combination doses (≥0.5 nmol/L: 5 μmol/L) were sufficient for clear synergy in MGMT low lines (NCI-H446 and NCI-H146). However, for the MGMT high lines (NCI-H82 and DMS114) a 4-fold increase in combination doses (≥ 2 nmol/L:20 μmol/L) was necessary for synergism (Fig. 5C and D; Supplementary Fig. S5A). This difference may be attributable to a requirement of achieving TMZ concentrations that exceed the demethylating capacity of MGMT. We confirmed similar findings with a lower talazoparib to TMZ ratio of 1:2,500, that coincided with the greatest excess over HSA, where synergy was seen in combination doses with at least 10 μmol/L of TMZ (Supplementary Fig. S5B and S5C).

Finally, we evaluated the combination of TMZ and talazoparib across the same seven PDX models in which we had established data on talazoparib single-agent activity. We hypothesized that this combination might extend the utility of PARP inhibition beyond the confines of SLFN11-high disease. We first determined a well-tolerated combination regimen of talazoparib 0.2 mg/kg daily and TMZ 6 mg/kg q4d (Supplementary Fig. S4A). We then conducted comparative assessment of each single agent versus the combination of TMZ and talazoparib across all seven models (Fig. 6A; Supplementary Fig. S6). We found that the combination exhibited significantly greater TGI than either single agent in multiple PDX models, although the combinatorial efficacy was particularly striking in high SLFN11-expressing models (i.e., JHU-LX22, JHU-LX110, SCRX-Lu149; Fig. 6A). Interestingly, in these models, MGMT expression levels did not appear to correlate with single-agent TMZ or combinatorial response (Fig. 6A and B). Notably, one MGMT-low PDX, JHU-LX48, exhibited dramatic sensitivity to TMZ, with marked tumor regression in all treated animals and with 2 of 5 animals in the TMZ treatment group and 3 of 5 animals in the combination treatment group having no tumor regrowth up to 180 days after the last TMZ dose was given. Other factors beyond the lack of MGMT expression are likely to contribute to TMZ efficacy in this exceptional responder. Collectively, these combinatorial treatment response data suggest that talazoparib and TMZ appear to be a highly promising combination therapy for SCLC.

PARP inhibition is a novel and promising therapy in SCLC, a recalcitrant disease where therapeutic advances have been limited. To date, the only response predictors associated with PARP inhibitor activity in cancer have been BRCA1/2 and ATM mutations in breast, ovarian, and prostate cancers. No such predictive biomarker has been defined for SCLC. Previous in vitro analyses in colon cancer and Ewing sarcoma cell lines have implicated SLFN11 as a correlate of response to conventional DNA damaging agents, including topoisomerase poisons and cisplatin (25, 44, 45). In this study, we substantially expand our knowledge of SLFN11 by showing for the first time that SLFN11 is a strong predictor of SCLC sensitivity to PARP inhibitors, and confirm these findings in vivo through multiple PDX models. In addition, we show that IHC staining against SLFN11 is a particularly strong predictor of PARP inhibitor response in PDXs, a finding that has immediate clinical translational implications. Finally, we show that combination therapy of a PARP inhibitor with TMZ is synergistic, well-tolerated and effective across multiple SCLC cell lines and PDXs.

Mu and colleagues recently presented an analysis of SLFN11 protein mechanisms of action, demonstrating direct interaction with replication protein A1 (RPA1), destabilization of RPA–ssDNA complexes, and ultimately inhibition of HR as observed in sister chromatid exchange and I-SceI florescent reporter gene conversion assays (21). These important data provide an evident mechanism for our series of observations, indicating that SLFN11 confers PARP inhibitor sensitivity in SCLC. Together, these studies suggest that high SLFN11 expression is a novel and potent mechanism of establishing a BRCA-like state of HRD, not necessarily reflected in current HRD scoring metrics, that governs sensitivity to PARP inhibitors and may dictate sensitivity to DNA damaging agents more generally.

We were surprised to find that all three HRD assays based on analysis of preexisting genomic scars failed to correlate with SCLC sensitivity to PARP inhibition. Our data suggest that SLFN11 upregulation could represent a more dynamic mechanism of HRD than germline BRCA gene deficiency and predicts PARP inhibitor sensitivity in SCLC without being evident in the analyses of genomic scarring which may take many cell generations to accumulate. At least in SCLC, SLFN11 appears to be a more predictive biomarker of PARP inhibitor efficacy than HRD assays, which may in part reflect the high false positive rates of HRD scoring algorithms (31). Functional HR assays, for example ex vivo IHC evaluation of RAD51 recombinase (RAD51) nuclear localization after DNA damage, may provide a method to integrate these findings, circumventing the low positive predictive value of the HRD genomic scar assays and allowing for further refinement of these molecular biomarkers in prospective clinical trials (16, 22, 46). Of note, alternatively, it is possible that SLFN11 confers a distinct pattern of genomic instability that is not detected by these aforementioned HRD scar assays and is worthy of future investigation.

The potential for dynamic regulation of SLFN11 expression suggests epigenetic modification as a potential therapeutic strategy. It has been shown that SLFN11 promoter hypermethylation correlates with resistance to platinum agents and that DNA methyltransferase (DNMT) chemical inhibition or genetic loss reverses the hypermethylated state, thereby resensitizing cells to platinum drugs through SLFN11 reexpression in vitro (47). As DNMT inhibitors are currently FDA approved for hematologic diseases, utilization of these epigenetic modifying drugs in overcoming SLFN11 promoter methylation-dependent resistance to DNA damaging agents represents an attractive avenue to explore in future studies.

One interesting finding from our study is the observation that relative PARP inhibitor sensitivity in SCLC correlated more strongly with trapping potency than was observed in any other tumor type. Our results suggest that in the context of SCLC, PARP trapping may be the governing mechanism of drug efficacy. We eagerly await the results of the early SCLC clinical trials with talazoparib, the most potent PARP trapper available, as well as potential studies with other PARP inhibitors to provide evidence for or against this preclinical observation. Our expectations for such studies are tempered by our emerging data suggesting SLFN11 suppression contributes to acquired resistance to cytotoxic chemotherapy in SCLC (48). These data suggest that recurrent/relapsed SCLC patients may exhibit lower expression levels of SLFN11 and imply that clinical trials examining PARP inhibitors as monotherapy in the second- or later-line settings may demonstrate suboptimal efficacy due to resistance conferred by loss of SLFN11. Notably, the broader activity in combination with TMZ may allow for greater benefit in the context of recurrent disease.

As combination therapies can increase disease control and may inhibit development of drug resistance, we explored the combination of talazoparib with TMZ, one of the few second-line agents for SCLC recognized in the National Comprehensive Cancer Network guidelines. We demonstrated in SCLC cell lines and PDX drug efficacy experiments PARP inhibitors synergize with TMZ in vitro and demonstrate strong combinatorial efficacy in vivo. These results are in line with encouraging results in other histologies (39–41, 49, 50). These promising results warrant future clinical investigation, and such trials are currently being planned. Future prospective clinical trials will ultimately be needed to validate these promising preclinical findings.

In conclusion, we demonstrate that the role of SLFN11-dependent drug sensitivity extends beyond conventional DNA damaging agents to a targeted agent, is operant in SCLC, and combination therapy of PARP inhibitors with TMZ is an opportune therapeutic strategy. Future work in our laboratory will focus on investigating strategies to manipulate SLFN11 expression in tumors to increase the therapeutic ratio of PARP inhibition and other DNA damaging agents. Additional work will be necessary to translate these promising findings into biomarker-informed clinical trials of PARP inhibitors in SCLC.

J.T. Poirier reports receiving a commercial research grant from Biomarin. C.M. Rudin is a consultant/advisory board member for Bristol-Myers Squibb and Medivation. No potential conflicts of interest were disclosed by the other authors.

Conception and design: B.H. Lok, E.E. Gardner, S.N. Powell, J.T. Poirier, C.M. Rudin

Development of methodology: B.H. Lok, E.E. Gardner, J.T. Poirier

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.H. Lok, E.E. Gardner, V.E. Schneeberger, P. Desmeules, N. Rekhtman, E.d. Stanchina, B.A. Teicher, N. Riaz, J.T. Poirier

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.H. Lok, E.E. Gardner, V.E. Schneeberger, A. Ni, N. Rekhtman, E.d. Stanchina, N. Riaz, S.N. Powell, J.T. Poirier, C.M. Rudin

Writing, review, and/or revision of the manuscript: B.H. Lok, E.E. Gardner, V.E. Schneeberger, A. Ni, P. Desmeules, N. Rekhtman, B.A. Teicher, J.T. Poirier, C.M. Rudin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.T. Poirier

Study supervision: S.N. Powell, C.M. Rudin

The authors thank Xiaodong Huang and the members of the Anti-Tumor Assessment Core Facility for their technical assistance; Afsar Barlas and Katia O. Manova-Todorova of the Molecular Cytology Core Facility for assistance with immunohistochemical staining; Hsiu-Yu Liu and Ralph J. Garippa of the RNAi Core Facility for assistance with custom shRNA constructs; Y. Jerry Shen and G. Karen Yu of Biomarin Pharmaceutical Inc. and all members of the Rudin laboratory for their helpful discussions.

This work was supported by BioMarin Pharmaceutical Inc., the National Cancer Institute (P30 CA008748, U54 OD020355-01, and R01 CA197936, to C.M. Rudin), LUNGevity, and Free to Breathe (J.T. Poirier), Conquer Cancer Foundation of ASCO, and Radiological Society of North America (RR1634, B.H. Lok). Additional research funding provided by the Van Andel Research Institute through the Van Andel Research Institute – Stand Up To Cancer Epigenetics Dream Team Translational Research Grant. Stand Up To Cancer is a program of the Entertainment Industry Foundation, administered by American Association for Cancer Research.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.