Probing three-dimensional collective cancer invasion with DIGME

Previous work has demonstrated heterogeneity within collectively invading packs of lung cancer cells, including leader and follower cells that cooperate to facilitate invasion into the microenvironment. In order to better characterize the genetic differences between these two distinct cell types, we performed RNA-seq on purified populations of leader and follower cells from the H1299 non-small cell lung cancer cell line. Through this analysis, we identified that leaders and followers each contain distinct mutational and gene expression profiles, despite originating from the same parental cell line. Importantly, we identified 17 point mutations found uniquely in leaders and 18 point mutations found uniquely in followers, thus representing the first known compilation of leader- and follower-enriched genetic variants. Notable leader-enriched mutated genes included NAE1, NUP93 and ACTR3, while notable follower-enriched mutated genes included NADK, NDUFS1 and LEO1, suggesting possible roles for these genes in the biological function of leader and follower cells. After validating these mutations in the genomic DNA of leaders and followers, we sought to determine whether these mutational signatures are correlated with leader and follower gene expression markers within individual cells in a heterogeneous tumor cell population. To this end, we performed single-cell RNA-seq on H1299 parental, leader and follower cells isolated directly from collectively invading packs in a 3-D matrix. Hierarchical clustering and tSNE analysis based upon most variably expressed genes revealed four distinct cell clusters; two expressing higher levels of leader cell genes including MYO10 and JAG1 (clusters 1 and 4), and two expressing higher levels of follower genes including IL13RA2 and HTATIP2 (clusters 2 and 3). Interestingly, clusters 1 and 4 are composed exclusively of cells with leader mutational profiles, while nearly all cells in clusters 2 and 3 contain follower mutational profiles, suggesting that these mutations could be driving the differential gene expression, and ultimately the unique biological behaviors, of leaders and followers. In addition, clusters 1 and 2 display higher levels of proliferative markers compared with clusters 3 and 4, an indication that there are proliferative and non-proliferative subpopulations within the leader and follower populations. Taken together, these data give us new insight into the multiple levels of heterogeneity that exist within an invasive tumor cell population, suggest novel drivers of leader and follower cell biology in collective invasion, and open the door to new potential strategies for targeting and inhibiting metastasis in human lung cancer. Citation Format: Brian A. Pedro, Jessica Konen, Emily Summerbell, Janna K. Mouw, Manali Rupji, Bhakti Dwivedi, Jeanne Kowalski, Paula M. Vertino, Adam I. Marcus. Dissecting the biology of leader and follower cells in collective cancer invasion [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 4590.

Much attention has been paid on the mechanism of cancer invasion from the viewpoint of the behaviour of individual cancer cells. On the other hand, histopathological analyses of specimens from cancer patients and of cancer invasion model animals have revealed that cancer cells often exhibit collective invasion, characterized by sustained cell-to-cell adhesion and polarized invasion as cell clusters. Interestingly, it has recently become evident that during collective invasion of cancer cells, the cells localized at invasion front (leader cells) and the cells following them (follower cells) exhibit distinct cellular characteristics, and that there exist the cells expressing representative proteins related to both epithelial and mesenchymal properties simultaneously, designated as hybrid epithelial-to-mesenchymal transition (EMT)-induced cells, in cancer tissue. Furthermore, the findings that cells adopted in hybrid EMT state form clusters and show collective invasion in vitro emphasize an importance of hybrid EMT-induced cells in collective cancer invasion. In this article, we overview recent findings of the mechanism underlying collective invasion of cancer cells and discuss the possibility of controlling cancer invasion and metastasis by targeting this process.

Background: Around 5% of all cutaneous squamous cell carcinoma (cSCC) metastasise. Metastases usually locate in regional skin and lymph nodes, suggesting collective cancer invasion. The cellular level of tumour invasion and prognostic parameters remain to be characterised. Methods: We performed immunohistochemical analyses of E-cadherin (marker for collective cancer invasion) and podoplanin (marker for epithelial-mesenchymal transition [EMT], single-cell invasion) expression in 102 samples of metastatic and non-metastatic cSCC and 18 corresponding skin and lymph node metastases to characterise the invasion of cSCC. Immunohistochemical results were retrospectively correlated with clinical data. Results: E-cadherin was highly expressed in metastatic and non-metastatic cSCC and skin metastases. This suggests collective cancer invasion. However, E-cadherin was downregulated in poorly differentiated cSCC and lymph node metastases, suggesting partial EMT. Podoplanin was significantly upregulated in metastatic (p = 0.002) and poorly differentiated (p = 0.003) cSCC. Overexpression of podoplanin represented a statistically independent prognostic factor for disease-free survival (p = 0.014). Conclusion: Collective cancer invasion is likely in cSCC. In lymph node metastases and poorly differentiated cSCC, partial EMT is possible. Podoplanin is an independent prognostic parameter for metastasis.

During tumor metastasis, cancer cells invade as a phenotypically heterogeneous collective pack to navigate the tumor microenvironment. Leader cells pioneer invasion into the microenvironment whereas follower cells immediately attach to and follow the leaders. To dissect the molecular mechanisms underlying this phenotypic heterogeneity, we developed a technique termed spatiotemporal genomic and cellular analysis (SaGA), which uses image-guided genomics to precisely select living, rare cell subpopulations that are maintained within a physiologically relevant environment for downstream genomic and molecular analyses. In this manner, we can precisely select, isolate, and amplify any living cell based upon phenotypic criteria. Using H1299 lung cancer cells, we precisely selected as few as 10 leader cells using the SaGA technique and extracted them from the bulk of a multicellular cancer spheroid embedded within a 3-D matrix. These isolated cells were used to create the first leader and follower purified cell cultures. The purified leader cell cultures continually show dynamic invasive patterns over time, while follower cells alone have limited invasive capabilities. Reintroducing limited numbers of leader cells, or even leader cell conditioned media, into follower cell cultures promoted invasion and motility in the follower cells. Gene expression analysis comparing leader versus follower cell lines showed significant enrichment in cell adhesion and vascular endothelial growth factor (VEGF) signaling pathways in leader cells. This was confirmed by protein analysis showing that leader cells secrete high levels of VEGF, and VEGF inhibition abolished leader-follower collective invasion, suggesting vascularity sprouting mimicry during chain formation. Additional analysis confirmed that leader cells utilize focal adhesion kinase-fibronectin signaling to create tension force to promote chain motility. While leader cells provide an escape mechanism for followers, follower cells in turn provide leaders with increased proliferation and survival. These data support a symbiotic model of collective invasion where different cell subtypes cooperate to promote successful escape. Overall, our data demonstrate that SaGA can precisely select living cells based upon dynamic behaviors for genomic analysis and can amplify rare cell populations for subsequent molecular, cellular, and proteomic analyses. Therefore, this image-guided method has the potential to impact the field of tumor heterogeneity by uncovering genomic signatures of rare yet dynamic subpopulations within a heterogeneous cancer population. Citation Format: Jessica Konen, Emily Summerbell, Kornelia Galior, Bhakti Dwivedi, Khalid Salaita, Jeanne Kowalski, Adam Marcus. Image-guided genomics reveals a symbiotic relationship between heterogeneous phenotypes in collective cancer invasion. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr LB-355.

Better understanding the mechanisms underlying the metastatic process is essential to developing novel targeted therapeutics. Recently, invadopodia have been increasingly recognized as important drivers of local invasion in metastasis. Invadopodia are actin-rich structures that have the ability to degrade the underlying extracellular matrix (ECM) and promote cell invasion. However, many questions concerning the mechanisms of invadopodia formation, regulation and function remain open to debate. Meanwhile, invasion of cells into the surrounding tissue and destruction of normal tissue architecture are two hallmarks of malignant tumors. Morphologically, two patterns of tumor invasion can be distinguished; single cell and collective cell invasion. The invasion of single cells and collective cells is often correlated with dramatic changes in the expression and function of adhesive and regulatory proteins. These changes are reminiscent of early developmental processes when cells acquire a migratory, mesenchymal phenotype. Currently, the role of invadopodia in collective cancer invasion is completely unknown. In the present study, we reveal that insuline-like growth factor-II mRNA-binding protein-3 (IMP-3) associated with invadopodial infiltration and collective cancer invasion through podoplanin (PDPN) mRNA stabilization as a component of invadopodia, suggesting that invadopodia plays a pivotal role in invasive front of collective cancer invasion and thereby involved the cancer metastasis. This finding, combined with other investigations into the mechanisms of invadopodia formation, reveal clinically-relevant targets for intervention in invadopodia, including IMP-3 and PDPN. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 4326. doi:1538-7445.AM2012-4326

Collective invasion, the coordinated movement of cohesive packs of cells, has become recognized as a major mode of metastasis for solid tumors. These packs are phenotypically heterogeneous and include specialized cells that lead the invasive pack and others that follow behind. To better understand how these unique cell types cooperate to facilitate collective invasion, we analyzed transcriptomic sequence variation between leader and follower populations isolated from the H1299 non-small cell lung cancer cell line using an image-guided selection technique. We now identify 14 expressed mutations that are selectively enriched in leader or follower cells, suggesting a novel link between genomic and phenotypic heterogeneity within a collectively invading tumor cell population. Functional characterization of two phenotype-specific candidate mutations showed that ARP3 enhances collective invasion by promoting the leader cell phenotype and that wild-type KDM5B suppresses chain-like cooperative behavior. These results demonstrate an important role for distinct genetic variants in establishing leader and follower phenotypes and highlight the necessity of maintaining a capacity for phenotypic plasticity during collective cancer invasion.

Tumor dissemination into the surrounding stroma is the initial step in a metastatic cascade. Invasion into stroma is a non-autonomous process for cancer, and its progression depends upon the stage of cancer, as well as the cells residing in the stroma. However, a systems framework to understand how stromal fibroblasts resist, collude, or aid cancer invasion has been lacking, limiting our understanding of the role of stromal biology in cancer metastasis. We and others have shown that gene perturbation in stromal fibroblasts can modulate cancer invasion into the stroma, highlighting the active role stroma plays in regulating its own invasion. However, cancer invasion into stroma is a complex higher-order process and consists of various sub-phenotypes that together can result in an invasion. Stromal invasion exhibits a diversity of modalities in vivo. It is not well understood if these diverse modalities are correlated, or they emanate from distinct mechanisms and if stromal biology could regulate these characteristics. These characteristics include the extent of invasion, formation, and persistence of invasive forks by cancer as opposed to a collective frontal invasion, the persistence of invading velocity by leader cells at the tip of invasive forks, etc. We posit that quantifying distinct aspects of collective invasion can provide useful suggestions about the plausible mechanisms regulating these processes, including whether the process is regulated by mechanics or by intercellular communication between stromal cells and cancer. Here, we have identified the sub-characteristics of invasion, which might be indicative of broader mechanisms regulating these processes, developed methods to quantify these metrics, and demonstrated that perturbation of stromal genes can modulate distinct aspects of collective invasion. Our results highlight that the genetic state of stromal fibroblasts can regulate complex phenomena involved in cancer dissemination and suggest that collective cancer invasion into stroma is an outcome of the complex interplay between cancer and stromal fibroblasts.

Collective migration is a crucial mechanism guiding cell movement in developmental processes and disease progression. Understanding the migration behavior of cell clusters is key to advancing our knowledge of morphogenesis, wound healing, and collective cancer invasion. Despite the understanding of the response of single cells to environmental physical cues, the collective behavior of cells in response to different levels of extracellular matrix stiffness is yet to be fully understood. Here, we present a quantitative investigation into how substrate stiffness and cell cluster size modulate the collective behavior and migration dynamics of NIH 3T3 fibroblasts. With the variation of PDMS and curing agent concentrations, two contrasting soft and stiff substrates with different stiffness were developed. Using a combination of atomic force microscopy (AFM) to precisely characterize substrate elastic moduli and time-lapse microscopy for tracking migration parameters, we demonstrate that substrate mechanics and cluster geometry synergistically govern collective behavior. Fibroblast migratory characteristics were greatly improved with increased stiffness and cluster size. Large clusters on stiff substrates exhibited greater circularity (~ 0.8), migration distance, displacement (135.6 µm), directionality (0.81), and velocity (24 µm/h) compared to single cells and small clusters on soft and stiff substrates. Moreover, detailed analysis of cytoskeletal reorganization via actin staining revealed the mechanotransductive pathways that convert physical cues into migratory behavior. These findings provide important insights into how substrate stiffness influences collective cell migration, offering potential applications in elucidating the mechanisms of morphogenesis and the dynamics of collective cell invasion during tumor progression.

Cancer cells collectively invade using a leader-follower organization, but the regulation of leader cells during this dynamic process is poorly understood. Using a dual double-stranded locked nucleic acid (LNA) nanobiosensor that tracks long noncoding RNA (lncRNA) dynamics in live single cells, we monitored the spatiotemporal distribution of lncRNA during collective cancer invasion. We show that the lncRNA MALAT1 (metastasis-associated lung adenocarcinoma transcript 1) is dynamically regulated in the invading fronts of cancer cells and patient-derived spheroids. MALAT1 transcripts exhibit distinct abundance, diffusivity, and distribution between leader and follower cells. MALAT1 expression increases when a cancer cell becomes a leader and decreases when the collective migration process stops. Transient knockdown of MALAT1 prevents the formation of leader cells and abolishes the invasion of cancer cells. Taken together, our single-cell analysis suggests that MALAT1 is dynamically regulated in leader cells during collective cancer invasion.

Collective invasion of cells plays a fundamental role in tissue growth, wound healing, immune response and cancer metastasis. This paper aimed to investigate cytokeratin-14 (CK14) expression and analyze its association with collective invasion in the invasive front of salivary adenoid cystic carcinoma (SACC) to uncover the role of collective invasion in SACC. Here, in the clinical data of 121 patients with SACC, the positive expression of CK14 was observed in 35/121(28.93%) of the invasive front of SACC. CK14 expression in the invasive front, local regional recurrence and distant metastasis were independent and significant prognostic factors in SACC patients. Then, we found that in an ex vivo 3D culture assay, CK14 siRNA receded the collective invasion, and in 2D monolayer culture, CK14 overexpression induced a collective SACC cell migration. These data indicated that the presence of characterized CK14+ cells in the invasive front of SACC promoted collective cell invasion of SACC and may be a biomarker of SACC with a worse prognosis.

Collective cancer invasion exhibits a hierarchical structure characterized by leader-follower organization. Dynamic gene expression analysis of invading cells using nanobiosensors within 3D microenvironments provides a valuable means to explore the regulation of leader cells during collective cancer invasion. Nonetheless, the analysis of time-lapse, multimodal images that capture the intricacies of complex invading structures and gene expression profiles in 3D tumor spheroids poses a significant technological challenge. Here, we present a computer vision-based workflow that streamlines the identification of protrusions and detached clusters from 3D tumor spheroids. This methodology not only discerns invading multicellular structures and quantifies their physical properties, but also captures gene expression patterns associated with these invasive mechanisms using an intracellular nanobiosensor. Consequently, it empowers a systematic exploration of the genotypic and phenotypic heterogeneities inherent in cancer invasion. To illustrate the effectiveness of this approach, we applied it to the analysis of a long noncoding RNA, MALAT1, in tumor spheroids derived from patients with muscle-invasive bladder cancer. Our investigation delved into the heterogeneity of cancer invasion and its relationship to MALAT1 expression. Overall, this workflow represents a valuable tool for gaining insights into the complexities of cancer invasion.

We developed a three-dimensional (3D) engineered model of a solid human breast tumor to study the effects of interstitial fluid pressure (IFP) on collective invasion and the expression levels of epithelial-mesenchymal transition (EMT) markers. Many solid tumors exhibit elevated IFP; as these tumors grow, intra-tumoral vascular and lymphatic vessels collapse. The non-functioning lymphatic system impairs drainage, and immature hyperpermeable blood vessels cause fluid to accumulate within the interstitial space. As a result, IFP rises steeply beyond the tumor periphery and plateaus at pressures as high as 50 mm Hg above normal at the tumor core. This pressure profile, in turn, leads to outward fluid flow from the core of the tumor. IFP has been shown to affect the migratory behavior of individual cells in 3D cell culture models, though its role in collective cancer invasion remains unknown. Moreover, the underlying molecular mechanisms linking IFP to changes in cell motility remain unclear. We sought to address these questions using our engineered model. Our 3D culture model consists of an aggregate of MDA-MB-231 breast cancer cells (mimicking a solid tumor) embedded within a 3D collagen gel that is flanked by two media reservoirs. The IFP profile experienced by the cancer cells is established by altering the heights of the media reservoirs on either side of the collagen, creating a hydrostatic pressure gradient. Transcript levels of EMT markers in the aggregates subjected to a variety of pressure profiles were determined using quantitative real-time PCR. Expression of these markers was also manipulated ectopically through the creation of stable cell lines. Time-lapse imaging and cell tracking were used to determine the persistence and motility of individual cells within the aggregates. We found that the direction of IFP-induced flow determines the invasive phenotype of tumor cells. Additionally, high expression levels of both mesenchymal (Snail1, vimentin) and epithelial (E-cadherin, keratin-8) markers were characteristic of collectively invading aggregates, suggesting that partial EMT is important for collective invasion. Ectopic expression and knockdown of EMT markers revealed that they are necessary and sufficient for collective invasion in response to IFP. Time-lapse imaging analysis demonstrated that IFP and EMT marker expression also affect the motility and persistence of individual cells within the aggregates, further confirming that IFP is an important regulator of collective invasion. In conclusion, we used a robust culture model of a human breast tumor to gain insight into the mechanisms guiding collective invasion from primary tumors in response to IFP; IFP alters expression levels of EMT markers, thereby regulating collective invasion. Citation Format: Alexandra S. Piotrowski-Daspit, Joe Tien, Celeste M. Nelson. Interstitial fluid pressure alters cell motility and collective invasion via EMT marker expression in an engineered model of a human breast tumor. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 4269.

Breast cancer undergoes collective tissue invasion and, in experimental models, can collectively metastasize. The prevalence of collective invasion and its contribution to distant metastasis in clinical disease, however, remains poorly defined. We here scored the adipose tissue invasion of primary invasive ductal carcinoma (IDC), expressing E-cadherin, and E-cadherin negative invasive lobular carcinoma (ILC) and identified predominantly collective invasion patterns (86/86 samples) in both carcinoma types. Whereas collective invasion in IDC lesions retained adherens junctions, multicellular clusters and “Indian files” in ILC, despite the absence of adherens junctions (AJ) proteins E-cadherin and β-catenin, retained CD44 at cell–cell contacts. By histomorphological scoring and semi-automated image analysis, we show that the extent of collective invasion into the adipose tissue correlated with decreased distant metastasis-free survival (5-year follow-up; hazard ratio: 2.32 and 2.29, respectively). Thus, collective invasion represents the predominant invasion mode in breast cancer, develops distinct junctional subtypes in IDC and ILC, and associates with distant metastasis, suggesting a critical role in systemic dissemination.

Colorectal cancer (CRC) is the third most commonly diagnosed cancer and the second most common cause of cancer death worldwide. The 5-year survival rate is 90% in patients with localized tumors, but the survival rate drastically decreases to 14% in patients with distant metastasis. Therefore, it is imperative to elucidate the molecular mechanisms underlying CRC metastasis. The RNA-binding protein LIN28B is overexpressed in over 30% of patients with CRC and is associated with poor prognosis. Indeed, our previous work revealed that LIN28B promotes liver metastasis in a subcutaneous xenograft model of CRC as well as a portal vein injection model. In the present study, we investigated the mechanism by which LIN28B promotes colorectal tumor progression and metastasis. To assess biological changes induced by LIN28B, we established CRC cells (DLD-1 and LoVo) with high expression of LIN28B (LIN28Bhigh). LIN28B overexpression upregulated Claudin-1 (CLDN1), a protein that functions as a major constituent of the tight junction complexes. RNA immunoprecipitation revealed that LIN28B directly binds to and stabilizes CLDN1 at a post-transcriptional level. Knockdown of CLDN1 expression in LIN28Bhigh cells suppresses cell aggregation, wound healing rate, and collective invasion. Using a mouse model of metastatic CRC, we reveal that knockdown of CLDN1 inhibits liver metastasis of LIN28Bhigh CRC cells. RNA-sequencing of metastatic liver tumors from LIN28Bhigh cells showed that NOTCH3 works downstream of the LIN28B-CLDN1 axis to induce cell aggregation, collective invasion, and metastatic liver tumor formation. Inhibition of Notch signaling by DAPT, an inhibitor of the γ-secretase complex, significantly reduced metastatic liver tumors in vivo. Taken together, our results indicate that LIN28B promotes cell aggregation, invasion, and liver metastasis in CRC through post-transcriptional induction of CLDN1 and upregulation of NOTCH3. Development of new therapies that target LIN28B-CLDN1-NOTCH3 axis may provide an effective strategy for stage 4 CRC. This work was supported by 9R01CA277795-22. Citation Format: Alice E. Shin, Kensuke Sugiura, Yasunori Masuike, Kensuke Suzuki, Christopher J. Lengner, Anil K. Rustgi. LIN28B promotes collective cell invasion and colorectal cancer metastasis via a novel CLDN1 and NOTCH3 axis [abstract]. In: Proceedings of the AACR Special Conference on Colorectal Cancer; 2022 Oct 1-4; Portland, OR. Philadelphia (PA): AACR; Cancer Res 2022;82(23 Suppl_1):Abstract nr B007.

Myosin-X (MYO10) is a noncanonical myosin that drives filopodia formation and extension, and is necessary for invasion and metastasis of breast cancer, prostate cancer and melanoma. It is not yet known by what mechanism MYO10 increases cancer cell invasion and metastasis. Two independently published MYO10 knockout mouse models showed severe developmental defects dependent upon collective migration of several cell types, suggesting that MYO10 may also regulate collective invasion of cancer cells. In order to study collective cancer cell invasion, our lab previously developed a technique to isolate and culture purified highly-invasive leader cells and highly-proliferative follower cells from collectively invading lung cancer cell lines. Using this method, we demonstrated distinct gene expression patterns, DNA methylation patterns, and phenotypic properties of leader versus follower cells. MYO10 is highly overexpressed in leader cells but is not expressed in follower cells, and MYO10 overexpression in leaders correlates with highly significant CpG island promoter DNA hypomethylation and gene body DNA hypermethylation. We hypothesize that MYO10 regulates cancer cell collective invasion by driving the formation of long filopodia necessary for leading-edge fibronectin patterning and subsequent leader cell invasion. MYO10 localizes at leader cell filopodia tips but not in follower cell filopodia within both 2D culture and 3D invading spheroids of four non-small cell lung cancer cell lines. MYO10 knockdown in purified leader cells and in these four lung cancer cell lines decreases the number and length of filopodia, 2D cell motility and 3D spheroid invasion; in contrast, MYO10 overexpression in follower cells increases filopodia length, cell motility and spheroid invasion. In addition, proteomic analysis shows that leader cells produce and secrete fibronectin (FN1), unlike follower cells. Immunofluorescence within invading spheroids shows the formation of long FN1 fibrils, i.e. elongated parallel bundles of FN1 that extend far past the leader cell body in multiple cell lines. FN1 fibrils preferentially localized with MYO10+ filopodia. Knockdown of MYO10 disrupts the formation of these fibronectin fibrils. We previously reported that knockdown of fibronectin completely abrogates lung cancer spheroid collective invasion. Therefore, our data suggest that MYO10 regulates cancer cell collective invasion by driving the formation of long filopodia that then regulate fibronectin architecture and subsequent invasion by leader cells. Note: This abstract was not presented at the meeting. Citation Format: Emily R. Summerbell, Jessica Konen, Jeanne Kowalski, Paula Vertino, Adam Marcus. MYO10 aberrant methylation and overexpression in leader cells regulates lung cancer collective cell invasion and fibronectin patterning [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 178.