Cell line authentication information

Supplementary Table S1. Detailed information about the antibodies used. Supplementary Table S2. Proportions of listed immune cell subsets in the tumor microenvironment of NSCLC GEMs were evaluated by FACS utilizing indicated lineage markers and proteins. Supplementary Table S3. Tumor cells from treated KP mice were subjected to RNA-sequencing. Supplementary Figure S1. Kinetics of tumor growth in KP mice treated with anti-PD-1. Supplementary Figure S2. Immune cell proportions in the tumors of treated KP mice. Supplementary Figure S3. PD-1 is prominently expressed on tumor-infiltrating T cells in GEMM of NSCLC. Supplementary Figure S4. Decreased expression of PD-1 and CTLA-4 on tumor-infiltrating CD4+ T cells under JQ1 treatment. Supplementary Figure S5. Increased expression of MHC class I on tumor-infiltrating tumor-associated macrophages (TAMs). Supplementary Figure S6. Combination of JQ1 and PD-1 blockade evokes unique genomic genomic changes in tumors of treated KP mice. Supplementary Figure S7. Overall survival of tumor-bearing nude mice treated with JQ1 or anti-PD-1. Supplementary Figure S8. JQ1 and/or anti-PD-1 treatment facilitated enhanced effector function of tumor-infiltrating CD8+ T cells. Supplementary Figure S9. Combination of JQ1 and α-PD-1 did not have substantial deleterious effects on tumor-associated T cells. Supplementary Figure S10. Tumor-infiltrating KLRG1+ CD4+Foxp3+ Tregs exhibit a highly activated profile but resemble natural peripheral Tregs. Supplementary Figure S11. Tumor-infiltrating KLRG1+ CD4+Foxp3+ Tregs are not derived from CD4+CD25- cells. Supplementary Figure S12. Tumor-associated KLRG1+ Tregs express higher CTLA-4 and demonstrate enhanced suppressive function. Supplementary Figure S13. Decreased expression of Treg signature proteins and Bcl-2 on tumor-infiltrating KLRG1+ Treg subset under JQ1 treatment.

Protein modulation by AZD5153 in hematological cell lines

<p>Figure S1: Downregulation of CCR2, CD180, MYC and upregulation of HEXIM1 mRNAs by AZD5153 in cells can be detected at 4h; Figure S2: Other BRD4/BET compounds can also downregulate CCR2 and CD180 mRNAs in hematological tumor cell lines; Supplementary Table 1: Description of cell lines used in the RNAseq study; Supplementary Table 2: Information on qPCR assays used in these studies; Supplementary Table 3: More detailed information for each healthy volunteer and patient sample.</p>

PURPOSE:To investigate protein and genetic status of BRAF V600E in ameloblastomas,ameloblastic carci-nomas and cysts,and to explore if recurrence and malignant transformation of ameloblastomas depends on BRAF status.METHODS:Twenty cysts,25 primary ameloblastomas,25 recurrent ameloblastomas and 8 ameloblastic carcinomas were analysed by immunohistochemistry,and 2 cysts,5 primary ameloblatomas,5 recurrent ameloblastomas,3 atypical ameloblastomas and 6 ameloblastic carcinomas were analysed by quantitative real-time PCR(qPCR).SPSS 26.0 software package was used for data analysis.RESULTS:There was no correlation between BRAF V600E expression and recur-rence.Cysts and carcinomas had lower frequencies of BRAF V600E mutations than ameloblastomas.There was a concor-dance between positive staining of BRAF V600E protein and BRAF V600E mutation.CONCLUSIONS:BRAF V600E mutation may be useful in the differential diagnosis of ameloblastomas with cysts and ameloblastic carcinomas.Immuno-histochemistry may be a beneficial method to screen for BRAF V600E mutation in ameloblastoma.

figure legends for supplementary Table S1-S3 and supplementary Figure S1-S13

Supplementary Figure 1. AZD5153 shows reduced BRD4 binding activity when BD2 function is abolished; Supplementary Figure 2. AZD5153 modulates MYC protein levels across hematologic cancer cell lines; Supplementary Figure 3. Gene Ontology (GO) analysis of the down-regulated transcripts by AZD5153 across all cell line types; Supplementary Figure 4. IC50 and percent maximal cell kill by AZD5153 and I-BET762 in five hematologic cell lines; Supplementary Figure 5. Tumor growth inhibition across six hematological xenograft Models

JAK1, JAK2, JAK3, and TYK2 belong to the JAK (Janus kinase) family. They play critical roles in cytokine signaling. Constitutive activation of JAK/STAT pathways is associated with a wide variety of diseases. Particularly, pSTAT3 is observed in response to the treatment with inhibitors of oncogenic signaling pathways such as EGFR, MAPK, and AKT and is associated with resistance or poorer response to agents targeting these pathways. Among the JAK family kinases, JAK1 has been shown to be the primary driver of STAT3 phosphorylation and signaling; therefore, selective JAK1 inhibition can be a viable means to overcome such treatment resistances. Herein, an account of the medicinal chemistry optimization from the promiscuous kinase screening hit 3 to the candidate drug 21 (AZD4205), a highly selective JAK1 kinase inhibitor, is reported. Compound 21 has good preclinical pharmacokinetics. Compound 21 displayed an enhanced antitumor activity in combination with an approved EGFR inhibitor, osimertinib, in a preclinical non-small-cell lung cancer (NSCLC) xenograft NCI-H1975 model.

Enhancer of zeste homologue 2 (EZH2), the catalytic subunit of polycomb repressive complex 2 (PRC2), regulates chromatin state and gene expression by methylating histone H3 lysine 27. EZH2 is overexpressed or mutated in various hematological malignancies and solid cancers. Our previous efforts to identify inhibitors of PRC2 methyltransferase activity by high-throughput screening (HTS) resulted in large numbers of false positives and thus a significant hit deconvolution challenge. More recently, others have reported compounds that bind to another PRC2 core subunit, EED, and allosterically inhibit EZH2 activity. This mechanism is particularly appealing as it appears to retain potency in cell lines that have acquired resistance to orthosteric EZH2 inhibition. By designing a fluorescence polarization probe based on the reported EED binding compounds, we were able to quickly and cleanly re-triage our previously challenging HTS hit list and identify novel allosteric PRC2 inhibitors.

Centrosomes function as the main microtubuleorganizing centers (MTOCs) and play an essential role during cellular mitosis. As such, centrosome number is tightly controlled, with centrosomes duplicated only once per cell cycle [1]. Unsurprisingly, centrosome abnormality, in particular centrosome amplification, has been frequently observed in human malignancies and is linked with aneuploidy and tumorigenesis. Recently the causative role of centrosome amplification in cancer has been demonstrated in a mouse model [2]. The resulting extra centrosomes from amplification pose a challenge for the proper execution of mitosis since multipolar mitosis is often detrimental to dividing cells and can lead to cell death [3]. Cancer cells, however, have developed tactics to avoid deleterious spindle multipolarity during mitosis through mechanisms such as centrosome clustering, a process that involves assembling multiple centrosomes into pseudo-bipolar spindles and thereby promotes tumor survival [4]. In recent years, a therapeutic strategy to pharmacologically impair centrosome clustering has emerged and demonstrated early promise in the preclinical setting: the goal is to exploit the vulnerability of cancer cells in handling spindle multipolarity by inducing centrosome de-clustering and achieve selective killing of cancer cells while sparing normal tissues. We recently reported the involvement of monoADP-ribosyltransferase PARP6 in maintaining centrosome integrity during mitosis by demonstrating that pharmacological inhibition of PARP6 induced multipolar spindle formation and elicited therapeutic effects in breast cancer [5]. This study was largely enabled by the AstraZeneca collection of PARP inhibitors consisting of mostly phthalazinone and quinazolinone based NAD+ mimetic cores that display differential selectivity towards polyand mono-ADP-ribosyltranferases [6]. Previous siRNA screens implicated PARP activity in centrosome clustering and bipolar spindle formation as a specific mitotic susceptibility in cancer cells [7]. We leveraged this knowledge and conducted parallel biochemical screens against multiple PARP enzymes together with cell-based phenotypic profiling in cancer cells to find that only PARP6 inhibition could drive multipolar spindle formation, with a strong correlation (R2 = 0.76) observed between PARP6 enzymatic inhibition and multipolar spindle induction potency. We recapitulated the mitotic phenotype with siRNA-based PARP6 knockdown. Together, these results link the suppression of PARP6 enzymatic activity to the induced multipolar spindle centrosome phenotype. We further showed that targeting PARP6 with the optimized small molecule inhibitor AZ0108 effectively induced spindle multipolarity and demonstrated promising anti-tumor activity in various breast cancer models in vitro and in vivo. To gain mechanistic understanding of the mitotic phenotype associated with PARP6 inhibition, we sought to identify PARP6 enzyme substrates using a powerful high-content ProtoArray screen consisting of 9,000+ full length human proteins. Centrosome related proteins were enriched as screening hits, including Chk1, which was further confirmed as a specific PARP6 substrate by in vitro enzymatic assays and mass spectrometric analysis in cells. Beyond its role in mediating DNA damage responses, Chk1 is also involved in regulating cellcycle checkpoints such as mitosis. During undisturbed cell cycle, Chk1 was reported to localize to interphase centrosomes and modulate cyclinB-Cdk1 activity, thereby impacting centrosomic mitotic events [8]. Our experiments demonstrated that intracellular ADP-ribosylation of Chk1 accumulated upon mitotic arrest and was abolished by treatment with PARP6 inhibitor AZ0108. Intriguingly, suppressing PARP6 using AZ0108 resulted in marked elevation of Chk1 phosphorylation at S345 and a dosedependent increase in centrosome-associated Chk1 expression. These events were accompanied by diminished activation of mitotic proteins such as FoxM1, histone H3 and Aurora kinases. Together, these observations support a novel mechanism for PARP6 in regulating Chk1 activation in mitosis via ADP-ribosylation and provide a mechanistic basis for multipolar spindle induction through PARP6 inhibition. Editorial

B-cell lymphoma 6 (BCL6) inhibition is a promising mechanism for treating hematological cancers but high quality chemical probes are necessary to evaluate its therapeutic potential. Here we report potent BCL6 inhibitors that demonstrate cellular target engagement and exhibit exquisite selectivity for BCL6 based on mass spectrometry analyses following chemical proteomic pull down. Importantly, a proteolysis-targeting chimera (PROTAC) was also developed and shown to significantly degrade BCL6 in a number of diffuse large B-cell lymphoma (DLBCL) cell lines, but neither BCL6 inhibition nor degradation selectively induced marked phenotypic response. To investigate, we monitored PROTAC directed BCL6 degradation in DLBCL OCI-Ly1 cells by immunofluorescence and discovered a residual BCL6 population. Analysis of subcellular fractions also showed incomplete BCL6 degradation in all fractions despite having measurable PROTAC concentrations, together providing a rationale for the weak antiproliferative response seen with both BCL6 inhibitor and degrader. In summary, we have developed potent and selective BCL6 inhibitors and a BCL6 PROTAC that effectively degraded BCL6, but both modalities failed to induce a significant phenotypic response in DLBCL despite achieving cellular concentrations.

Abstract KRAS mutation is present in approximately 30% of human lung adenocarcinomas. Although recent advances in targeted therapy have shown great promise, effective targeting of KRAS remains elusive, and concurrent alterations in tumor suppressors render KRAS-mutant tumors even more resistant to existing therapies. Contributing to the refractoriness of KRAS-mutant tumors are immunosuppressive mechanisms, such as increased presence of suppressive regulatory T cells (Treg) in tumors and elevated expression of the inhibitory receptor PD-1 on tumor-infiltrating T cells. Treatment with BET bromodomain inhibitors is beneficial for hematologic malignancies, and they have Treg-disruptive effects in a non–small cell lung cancer (NSCLC) model. Targeting PD-1–inhibitory signals through PD-1 antibody blockade also has substantial therapeutic impact in lung cancer, although these outcomes are limited to a minority of patients. We hypothesized that the BET bromodomain inhibitor JQ1 would synergize with PD-1 blockade to promote a robust antitumor response in lung cancer. In the present study, using Kras+/LSL-G12D; Trp53L/L (KP) mouse models of NSCLC, we identified cooperative effects between JQ1 and PD-1 antibody. The numbers of tumor-infiltrating Tregs were reduced and activation of tumor-infiltrating T cells, which had a T-helper type 1 (Th1) cytokine profile, was enhanced, underlying their improved effector function. Furthermore, lung tumor–bearing mice treated with this combination showed robust and long-lasting antitumor responses compared with either agent alone, culminating in substantial improvement in the overall survival of treated mice. Thus, combining BET bromodomain inhibition with immune checkpoint blockade offers a promising therapeutic approach for solid malignancies such as lung adenocarcinoma. Cancer Immunol Res; 6(10); 1234–45. ©2018 AACR.

34 on July 6, 2021. © 2018 American Association for Cancer Research. cancerimmunolres.aacrjournals.org Downloaded from Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on August 7, 2018; DOI: 10.1158/2326-6066.CIR-18-0077

Abstract KRAS mutation is present in about 30% of human lung adenocarcinomas. While recent advances in targeted therapy have shown great promise, KRAS remains undruggable and concurrent alterations in tumor suppressors render KRAS mutant tumors even more resistant to existing therapies. Contributing to the refractoriness of KRAS mutant tumors harboring these co-mutations are immunosuppressive mechanisms such as increased presence of suppressive Tregs in tumors and elevated expression of the inhibitory receptor PD-1 on tumor-infiltrating T cells. BET bromodomain inhibitors demonstrate clinical benefit in hematologic malignancies, and prior reports demonstrate their Treg-disruptive effects in a NSCLC model. Targeting PD-1 inhibitory signals through anti-PD-1 antibody blockade has also shown substantial therapeutic impact in lung cancer although these outcomes are still limited to a minor pool of patients. We therefore hypothesized that the BET bromodomain inhibitor JQ1 would synergize with PD-1 blockade to promote robust anti-tumor response in lung cancer. In the present study, using Kras+/LSL-G12D; Trp53L/L (KP) mouse models of non-small cell lung cancer, we identified cooperative effects among JQ1 and anti-PD-1 that included reduced numbers of tumor-infiltrated Tregs and enhanced activation of tumor-infiltrating T cells, which exhibited a Th1 cytokine profile that favored their demonstrated improved effector function. Furthermore, lung-tumor-bearing mice under this combinatorial treatment regimen showed robust and long-lasting anti-tumor responses compared to either agent alone, culminating in substantial improvement in the survival of treated mice. Thus, combining BET bromodomain inhibition with immune checkpoint blockade offers a promising therapeutic approach for solid malignancies such as lung adenocarcinoma. Citation Format: Dennis O. Adeegbe, Shengwu Liu, Michaela Bowden, Peter S. Hammerman, James E. Bradner, Anil K. Rustgi, Adam J. Bass, Gordon J. Freeman, Huawei Chen, kwok-Kin Wong. BET bromodomain inhibition synergizes with PD-1 blockade to facilitate anti-tumor response in Kras-mutant non-small cell lung cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 4713.

BRD4 is a transcriptional co-activator functioning to recruit regulatory complexes to acetylated chromatin. A subset of High-grade Serous Ovarian Cancer (HGSOC) patients are typified by focal, recurrent BRD4 gene amplifications. Despite previously described cancer dependencies, it is unclear whether BRD4 amplification events are oncogenic in HGSOC. We find that physiologically relevant levels of expression of BRD4 isoforms in non-transformed ovarian cells result in cellular transformation. Transcriptional profiling of BRD4-transformed ovarian cells, and BRD4-amplified HGSOC patient samples revealed shared expression patterns, including enriched MYC, and E2F1 gene signatures. Furthermore, we demonstrate that a novel BET inhibitor, AZD5153, is highly active in BRD4-amplified patient derived xenografts and uncover Neuregulin-1 as a novel BRD4 effector. Experiments involving Neuregulin-1 inhibition and exogenous addition, demonstrate Neuregulin-1 as necessary and sufficient for BRD4-mediated transformation. This study demonstrates the oncogenic potential of BRD4 amplification in cancer and establishes BRD4-amplified HGSOC as a potential patient population that could benefit from BET inhibitors.

Poly(ADP-ribose) polymerase inhibitors (PARPi) are selectively active in cells with homologous recombination (HR) deficiency (HRD) caused by mutations in BRCA1, BRCA2, and other pathway members. We sought small molecules that induce HRD in HR-competent cells to induce synthetic lethality with PARPi and extend the utility of PARPi. We demonstrated that inhibition of bromodomain containing 4 (BRD4) induced HRD and sensitized cells across multiple tumor lineages to PARPi regardless of BRCA1/2, TP53, RAS, or BRAF mutation status through depletion of the DNA double-stand break resection protein CtIP (C-terminal binding protein interacting protein). Importantly, BRD4 inhibitor (BRD4i) treatment reversed multiple mechanisms of resistance to PARPi. Furthermore, PARPi and BRD4i are synergistic in multiple in vivo models.

Previous reports have demonstrated that select cancers depend on BRD4 to regulate oncogenic gene transcriptional programs. Here we describe a novel role for BRD4 in DNA damage response (DDR). BRD4 associates with and regulates the function of pre-replication factor CDC6 and plays an indispensable part in DNA replication checkpoint signaling. Inhibition of BRD4 by JQ1 or AZD5153 resulted in a rapid, time-dependent reduction in CHK1 phosphorylation and aberrant DNA replication re-initiation. Furthermore, BRD4 inhibition sensitized cancer cells to various replication stress-inducing agents, and synergized with ATR inhibitor AZD6738 to induce cell killing across a number of cancer cell lines. The synergistic interaction between AZD5153 and AZD6738 is translatable to in vivo ovarian cell-line and patient-derived xenograft models. Taken together, our study uncovers a new biological function of BRD4 and provides mechanistic rationale for combining BET inhibitors with DDR-targeted agents for cancer therapy.

AbstractPARP proteins represent a class of post-translational modification enzymes with diverse cellular functions. Targeting PARPs has proven to be efficacious clinically, but exploration of the therapeutic potential of PARP inhibition has been limited to targeting poly(ADP-ribose) generating PARP, including PARP1/2/3 and tankyrases. The cancer-related functions of mono(ADP-ribose) generating PARP, including PARP6, remain largely uncharacterized. Here, we report a novel therapeutic strategy targeting PARP6 using the first reported PARP6 inhibitors. By screening a collection of PARP compounds for their ability to induce mitotic defects, we uncovered a robust correlation between PARP6 inhibition and induction of multipolar spindle (MPS) formation, which was phenocopied by PARP6 knockdown. Treatment with AZ0108, a PARP6 inhibitor with a favorable pharmacokinetic profile, potently induced the MPS phenotype, leading to apoptosis in a subset of breast cancer cells in vitro and antitumor effects in vivo. In addition, Chk1 was identified as a specific substrate of PARP6 and was further confirmed by enzymatic assays and by mass spectrometry. Furthermore, when modification of Chk1 was inhibited with AZ0108 in breast cancer cells, we observed marked upregulation of p-S345 Chk1 accompanied by defects in mitotic signaling. Together, these results establish proof-of-concept antitumor efficacy through PARP6 inhibition and highlight a novel function of PARP6 in maintaining centrosome integrity via direct ADP-ribosylation of Chk1 and modulation of its activity.Significance:These findings describe a new inhibitor of PARP6 and identify a novel function of PARP6 in regulating activation of Chk1 in breast cancer cells.

The BET (bromodomain and extra terminal domain) family proteins function as chromatin readers through interactions with acetylated lysine residues on both histones and transcription factors. BRD4 is the most well-studied member of this group that promotes tumorigenesis by regulating expression of core transcriptional programs; however, the role of Brd4 in other cellular processes is not entirely clear. Here, we show using small molecule inhibition of the BET proteins that Brd4 is required to maintain S-phase checkpoint signaling. Treatment of high-grade serous ovarian and osteosarcoma cancer cell lines with BET inhibitors, AZD5153 and JQ1, results in the accumulation of gH2AX and 53BP1 foci formation, which is suggestive of elevated DNA damage levels. The induction of DNA damage caused by both inhibitors corresponds with cell cycle arrest at the G1/S boundary. Displacement of Brd4 in cell lines under intrinsic or exogenous replication stress led to a time-dependent reduction in phospho-Chk1 first detected within 30 minutes of AZD5153 treatment in U2OS cells. The decrease in Chk1 phosphorylation was observed without a concomitant decrease in total Chk1, and this was not replicated by treatment with pan-transcriptional inhibitors, suggesting a non-transcriptional mechanism linked to Brd4. Furthermore, Brd4 interacts with core DNA replication machinery during S-phase and inhibition of Brd4 leads to aberrant DNA replication without decreasing ATR phosphorylation. Consistent with a role in S-phase signaling, Brd4 inhibition sensitizes replicating cells to replication stress-inducing agents. Finally, we observed that ovarian cancer cell lines are highly-sensitive to the combination of Brd4 and ATR (AZD6738) inhibitors. Concurrent treatment with AZD5153 and AZD6738 induces DNA damage and apoptosis in a synergistic manner. Together, these findings highlight a novel function for Brd4 and provide evidence to support further evaluation of Brd4i combinations with DNA damage repair agents. Citation Format: Jingwen Zhang, Huawei Chen, Austin Dulak. Brd4 modulates S-phase checkpoint response and sensitizes ovarian cancer cells to ATR inhibition. [abstract]. In: Proceedings of the AACR Precision Medicine Series: Targeting the Vulnerabilities of Cancer; May 16-19, 2016; Miami, FL. Philadelphia (PA): AACR; Clin Cancer Res 2017;23(1_Suppl):Abstract nr B21.

Janus kinases are a family of four enzymes; JAK1, JAK2, JAK3 and tyrosine kinase 2 (TYK2) that are critical in cytokine signalling, with constitutive activation of JAK/STAT pathways associated with a wide variety of malignancies. Elevated JAK/STAT signalling leading to increased activation of STAT3 is reported in a wide variety of cancers, including breast, liver, prostate, colorectal, head and neck, oesophageal, pancreatic, bladder, and non-small cell lung, and is implicated in the pathogenesis of diffuse large B-cell lymphoma and nasopharyngeal carcinomas. Overall, up to 70% of human tumours are linked to persistent elevated STAT3 activity which can be associated with poorer prognosis in many of these settings. In addition, elevated pSTAT3 is observed in response to chemotherapy treatment, and also in response to treatment with inhibitors of oncogenic signalling pathways such as EGFR, MAPK and AKT, and is associated with resistance or poorer response to agents targeting these pathways. In many of these cases, JAK1 is believed to be a primary driver of STAT3 phosphorylation and signalling, suggesting inhibition of JAKs as a therapeutic approach to treat these potential resistance mechanisms. The mixed JAK1/2 kinase inhibitor ruxolitinib is approved for the treatment of myeloproliferative neoplasms including intermediate or high risk myelofibrosis and polycythemia vera and has been tested in a variety of tumor settings. Since JAK2 is essential for the signal transduction downstream of erythropoietin, thrombopoietin and related receptors that control erythrocyte and megakaryocyte expansion, dosing of inhibitors that target JAK2 can be limited by toxicities such as thrombocytopenia and anaemia. Starting from a non-kinome selective screening hit, structure-based design was used to optimise a series of aminopyrimidines that led to JAK1-selective candidate drug AZD4205. This compound demonstrates ATP competitive binding with IC 50 ’s in a high ATP concentration enzyme assay against JAK1 of 73 nM (Ki = 2.8 nM), with high selectivity against JAK2 and JAK3 with IC 50 ’s of 13,233 nM and u003e30,000 nM respectively. In addition it showed potent inhibition of p-STAT3 in a cell based assay of JAK1 activity with an IC 50 of 128 nM and excellent selectivity across the kinome. In summary, AZD4205 is a highly potent JAK1-selective kinase inhibitor with excellent preclinical pharmacokinetics with potential for further clinical development. The optimization from screening hit to first disclosure of this candidate drug will be presented. Citation Format: Jason G. Kettle, Qibin Su, Neil Grimster, Sameer Kawatkar, Scott Throner, Richard Woessner, Huawei Chen, Geraldine Bebernitz, Kristen Bell, Erica Anderson, Linette Ruston, Jon Winter-Holt, Paul Lyne, Melissa Vasbinder, Claudio Chuaqui. Discovery of the JAK1 selective kinase inhibitor AZD4205 [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 979. doi:10.1158/1538-7445.AM2017-979