Predicting benefit from immune checkpoint inhibitors in patients with non-small-cell lung cancer by CT-based ensemble deep learning: a retrospective study.

BACKGROUND: Only around 20-30% of patients with non-small-cell lung cancer (NCSLC) have durable benefit from immune-checkpoint inhibitors. Although tissue-based biomarkers (eg, PD-L1) are limited by suboptimal performance, tissue availability, and tumour heterogeneity, radiographic images might holistically capture the underlying cancer biology. We aimed to investigate the application of deep learning on chest CT scans to derive an imaging signature of response to immune checkpoint inhibitors and evaluate its added value in the clinical context. METHODS: In this retrospective modelling study, 976 patients with metastatic, EGFR/ALK negative NSCLC treated with immune checkpoint inhibitors at MD Anderson and Stanford were enrolled from Jan 1, 2014, to Feb 29, 2020. We built and tested an ensemble deep learning model on pretreatment CTs (Deep-CT) to predict overall survival and progression-free survival after treatment with immune checkpoint inhibitors. We also evaluated the added predictive value of the Deep-CT model in the context of existing clinicopathological and radiological metrics. FINDINGS: Our Deep-CT model demonstrated robust stratification of patient survival of the MD Anderson testing set, which was validated in the external Stanford set. The performance of the Deep-CT model remained significant on subgroup analyses stratified by PD-L1, histology, age, sex, and race. In univariate analysis, Deep-CT outperformed the conventional risk factors, including histology, smoking status, and PD-L1 expression, and remained an independent predictor after multivariate adjustment. Integrating the Deep-CT model with conventional risk factors demonstrated significantly improved prediction performance, with overall survival C-index increases from 0·70 (clinical model) to 0·75 (composite model) during testing. On the other hand, the deep learning risk scores correlated with some radiomics features, but radiomics alone could not reach the performance level of deep learning, indicating that the deep learning model effectively captured additional imaging patterns beyond known radiomics features. INTERPRETATION: This proof-of-concept study shows that automated profiling of radiographic scans through deep learning can provide orthogonal information independent of existing clinicopathological biomarkers, bringing the goal of precision immunotherapy for patients with NSCLC closer. FUNDING: National Institutes of Health, Mark Foundation Damon Runyon Foundation Physician Scientist Award, MD Anderson Strategic Initiative Development Program, MD Anderson Lung Moon Shot Program, Andrea Mugnaini, and Edward L C Smith.

The Combiome Hypothesis: Selecting Optimal Treatment for Cancer Patients.

Existing approaches for cancer diagnosis are inefficient in the use of diagnostic tissue, and decision-making is often sequential, typically resulting in delayed treatment initiation. Future diagnostic testing needs to be faster and optimize increasingly complex treatment decisions. We envision a future where comprehensive testing is routine. Our approach, termed the "combiome," combines holistic information from the tumor, and the patient's immune system. The combiome model proposed here advocates synchronized up-front testing with a panel of sensitive assays, revealing a more complete understanding of the patient phenotype and improved targeting and sequencing of treatments. Development and eventual adoption of the combiome model for diagnostic testing may provide better outcomes for all cancer patients, but will require significant changes in workflows, technology, regulations, and administration. In this review, we discuss the current and future testing landscape, targeting of personalized treatments, and technological and regulatory advances necessary to achieve the combiome.

From clinical specimens to human cancer preclinical models-a journey the NCI-cell line database-25 years later.

The intramural the National Cancer Institute (NCI) and more recently the University of Texas Southwestern Medical Center with many different collaborators comprised a complex, multi-disciplinary team that collaborated to generated large, comprehensively annotated, cell-line related research resources which includes associated clinical, and molecular characterization data. This material has been shared in an anonymized fashion to accelerate progress in overcoming lung cancer, the leading cause of cancer death across the world. However, this cell line collection also includes a range of other cancers derived from patient-donated specimens that have been remarkably valuable for other types of cancer and disease research. A comprehensive analysis conducted by the NCI Center for Research Strategy of the 278 cell lines reported in the original Journal of Cellular Biochemistry Supplement, documents that these cell lines and related products have since been used in more than 14 000 grants, and 33 207 published scientific reports. This has resulted in over 1.2 million citations using at least one cell line. Many publications involve the use of more than one cell line, to understand the value of the resource collectively rather than individually; this method has resulted in 2.9 million citations. In addition, these cell lines have been linked to 422 clinical trials and cited by 4700 patents through publications. For lung cancer alone, the cell lines have been used in the research cited in the development of over 70 National Comprehensive Cancer Network clinical guidelines. Finally, it must be underscored again, that patient altruism enabled the availability of this invaluable research resource.

Versatile ion S5XL sequencer for targeted next generation sequencing of solid tumors in a clinical laboratory.

BACKGROUND: Next generation sequencing based tumor tissue genotyping involves complex workflow and a relatively longer turnaround time. Semiconductor based next generation platforms varied from low throughput Ion PGM to high throughput Ion Proton and Ion S5XL sequencer. In this study, we compared Ion PGM and Ion Proton, with a new Ion S5XL NGS system for workflow scalability, analytical sensitivity and specificity, turnaround time and sequencing performance in a clinical laboratory. METHODS: Eighteen solid tumor samples positive for various mutations as detected previously by Ion PGM and Ion Proton were selected for study. Libraries were prepared using DNA (range10-40ng) from micro-dissected formalin-fixed, paraffin-embedded (FFPE) specimens using the Ion Ampliseq Library Kit 2.0 for comprehensive cancer (CCP), oncomine comprehensive cancer (OCP) and cancer hotspot panel v2 (CHPv2) panel as per manufacturer's instructions. The CHPv2 were sequenced using Ion PGM whereas CCP and OCP were sequenced using Ion Proton respectively. All the three libraries were further sequenced individually (S540) or multiplexed (S530) using Ion S5XL. For S5XL, Ion chef was used to automate template preparation, enrichment of ion spheres and chip loading. Data analysis was performed using Torrent Suite 4.6 software on board S5XL and Ion Reporter. A limit of detection and reproducibility studies was performed using serially diluted DLD1 cell line. RESULTS: A total of 241 variant calls (235 single nucleotide variants and 6 indels) expected in the studied cohort were successfully detected by S5XL with 100% and 97% concordance with Ion PGM and Proton, respectively. Sequencing run time was reduced from 4.5 to 2.5 hours with output range of 3-5 GB (S530) and 8-9.3Gb (S540). Data analysis time for the Ion S5XL is faster 1 h (S520), 2.5 h (S530) and 5 h (S540) chip, respectively as compared to the Ion PGM (3.5-5 h) and Ion Proton (8h). A limit detection of 5% allelic frequency was established along with high inter-run reproducibility. CONCLUSION: Ion S5XL system simplified workflow in a clinical laboratory, was feasible for running smaller and larger panels on the same instrument, had a shorter turnaround time, and showed good concordance for variant calls with similar sensitivity and reproducibility as the Ion PGM and Proton.

Copper transporter CTR1 expression and tissue platinum concentration in non-small cell lung cancer.

BACKGROUND: Platinum resistance is a major limitation in the treatment of advanced non-small cell lung cancer (NSCLC). We previously demonstrated that low tissue platinum concentration in NSCLC specimens was significantly associated with reduced tumor response. Furthermore, low expression of the copper transporter CTR1, a transporter of platinum uptake was associated with poor clinical outcome following platinum-based therapy in NSCLC patients. We investigated the relationship between tissue platinum concentrations and CTR1 expression in NSCLC specimens. METHODS: We identified paraffin-embedded NSCLC tissue blocks of known tissue platinum concentrations from 30 patients who underwent neoadjuvant platinum-based chemotherapy at MD Anderson Cancer Center. Expression of CTR1 in tumors and normal adjacent lung specimens was determined by immunohistochemistry with adequate controls. RESULTS: Tissue platinum concentration significantly correlated with tumor response in 30 patients who received neoadjuvant platinum-based chemotherapy (P<0.001). CTR1 was differentially expressed in NSCLC tumors. A subset of patients with undetectable CTR1 expression in their tumors had reduced platinum concentrations (P=0.058) and tumor response (P=0.016) compared to those with any level of CTR1 expression. We also observed that African Americans had significantly reduced CTR1 expression scores (P=0.001), tissue platinum concentrations (P=0.009) and tumor shrinkage (P=0.016) compared to Caucasians. CONCLUSIONS: To our best knowledge this is the first study investigating the function of CTR1 in clinical specimens. CTR1 expression may be necessary for therapeutic efficacy of platinum drugs, consistent with previous preclinical studies. A prospective clinical trial is necessary to develop CTR1 into a potential biomarker for platinum drugs.

HER2YVMA drives rapid development of adenosquamous lung tumors in mice that are sensitive to BIBW2992 and rapamycin combination therapy.

Mutations in the HER2 kinase domain have been identified in human clinical lung cancer specimens. Here we demonstrate that inducible expression of the most common HER2 mutant (HER2(YVMA)) in mouse lung epithelium causes invasive adenosquamous carcinomas restricted to proximal and distal bronchioles. Continuous expression of HER2(YVMA) is essential for tumor maintenance, suggesting a key role for HER2 in lung adenosquamous tumorigenesis. Preclinical studies assessing the in vivo effect of erlotinib, trastuzumab, BIBW2992, and/or rapamycin on HER2(YVMA) transgenic mice or H1781 xenografts with documented tumor burden revealed that the combination of BIBW2992 and rapamycin is the most effective treatment paradigm causing significant tumor shrinkage. Immunohistochemical analysis of lung tumors treated with BIBW2992 and rapamycin combination revealed decreased phosphorylation levels for proteins in both upstream and downstream arms of MAPK and Akt/mTOR signaling axes, indicating inhibition of these pathways. Based on these findings, clinical testing of the BIBW2992/rapamycin combination in non-small cell lung cancer patients with tumors expressing HER2 mutations is warranted.

Hypermethylation of the death-associated protein kinase promoter attenuates the sensitivity to TRAIL-induced apoptosis in human non-small cell lung cancer cells.

Death-associated protein (DAP) kinase plays an important role in IFN-gamma, tumor necrosis factor (TNF)-alpha, or Fas-ligand induced apoptosis. TNF-related apoptosis-inducing ligand (TRAIL) is a member of the TNF ligand family and can induce caspase-dependent apoptosis in cancer cells while sparing most of the normal cells. However, some of the cancer cell lines are insensitive to TRAIL, and such resistance cannot be explained by the dysfunction of TRAIL receptors or their known downstream targets. We reported previously that DAP kinase promoter is frequently methylated in non-small cell lung cancer (NSCLC), and such methylation is associated with a poor clinical outcome. To determine whether DAP kinase promoter methylation contributes to TRAIL resistance in NSCLC cells, we measured DAP kinase promoter methylation and its gene expression status in 11 NSCLC cell lines and correlated the methylation/expression status with the sensitivity of cells to TRAIL. Of the 11 cell lines, 1 had a completely methylated DAP kinase promoter and no detectable DAP kinase expression, 4 exhibited partial promoter methylation and substantially decreased gene expression, and the other 6 cell lines showed no methylation in the promoter and normal DAP kinase expression. Therefore, the amount of DAP kinase expression amount was negatively correlated to its promoter methylation (r = -0.77; P = 0.003). Interestingly, the cell lines without the DAP kinase promoter methylation underwent substantial apoptosis even in the low doses of TRAIL, whereas those with DAP kinase promoter methylation were resistant to the treatment. The resistance to TRAIL was reciprocally correlated to DAP kinase expression in 10 of the 11 cell lines at 10 ng/mL concentration (r = 0.91; P = 0.001). We treated cells resistant to TRAIL with 5-aza-2'-deoxycytidine, a demethylating reagent, and found that these cells expressed DAP kinase and became sensitive to TRAIL. These results suggest that DAP kinase is involved in TRAIL-mediated cell apoptosis and that a demethylating agent may have a role in enhancing TRAIL-mediated apoptosis in some NSCLC cells by reactivation of DAP kinase.