Molecular modifiers of KRASG12C inhibitor (KRASG12Ci) efficacy in advanced KRASG12C-mutant NSCLC are poorly defined. In a large unbiased clinicogenomic analysis of 424 patients with non–small cell lung cancer (NSCLC), we identified and validated coalterations in KEAP1, SMARCA4, and CDKN2A as major independent determinants of inferior clinical outcomes with KRASG12Ci monotherapy. Collectively, comutations in these three tumor suppressor genes segregated patients into distinct prognostic subgroups and captured ∼50% of those with early disease progression (progression-free survival ≤3 months) with KRASG12Ci. Pathway-level integration of less prevalent coalterations in functionally related genes nominated PI3K/AKT/MTOR pathway and additional baseline RAS gene alterations, including amplifications, as candidate drivers of inferior outcomes with KRASG12Ci, and revealed a possible association between defective DNA damage response/repair and improved KRASG12Ci efficacy. Our findings propose a framework for patient stratification and clinical outcome prediction in KRASG12C-mutant NSCLC that can inform rational selection and appropriate tailoring of emerging combination therapies. Significance: In this work, we identify co-occurring genomic alterations in KEAP1, SMARCA4, and CDKN2A as independent determinants of poor clinical outcomes with KRASG12Ci monotherapy in advanced NSCLC, and we propose a framework for patient stratification and treatment personalization based on the comutational status of individual tumors. See related commentary by Heng et al., p. 1513. This article is highlighted in the In This Issue feature, p. 1501

Tumor mutational burden (TMB) is a new biomarker for prediction of response to PD-(L)1 treatment. Comprehensive sequencing approaches (i.e., whole exome and whole genome sequencing) are ideally suited to measure TMB directly. However, as their applicability in routine diagnostics is currently limited by high costs, long turnaround times and poor availability of fresh tissue, targeted next-generation sequencing (NGS) of formalin-fixed and paraffin-embedded (FFPE) samples appears to be a more feasible and straight-forward approach for TMB approximation, which can be seamlessly integrated in already existing diagnostic workflows and pipelines. In this work, we provide an overview of the clinical implications of TMB testing and highlight key parameters including pre-analysis, analysis and post-analytical steps that influence and shape TMB approximation by panel sequencing. Collectively, the data will not only serve as a field guide and state of the art knowledge source for molecular pathologists who consider implementation of TMB measurement in their lab, but also enable clinicians in understanding the specific parameters influencing TMB test results and reporting.

Panel-based estimates of Tumor Mutational Burden (psTMB) are increasingly replacing whole-exome-sequencing (WES) Tumor mutational burden as predictive biomarker of immune checkpoint blockade (ICB). A mathematical law describing psTMB variability was derived using a random mutation model and complemented by the contributions of non-randomly mutated real-world cancer genomes and intratumoral heterogeneity through simulations in publicly available datasets. The coefficient of variation (CV) of psTMB decreased inversely proportional with the square root of the panel size and the square root of the TMB level. In silico simulations of all major commercially available panels in the TCGA pan-cancer cohort confirmed the validity of this mathematical law and demonstrated that the CV was 35% for TMB=10 muts/Mbp for the largest panels of size 1.1-1.4 Mbp. Accordingly, misclassification rates (gold standard: WES) to separate “TMBhigh” from “TMBlow” using a cut-point of 199 mutations were 10-12% in TCGA-LUAD and 17-19% in TCGA-LUSC. A novel three-tier psTMB classification scheme which accounts for the likelihood of misclassification is proposed. Analysis of two independent datasets revealed that small gene panels were poor predictors of ICB response. Moreover, we noted significant intratumoral variance of psTMB scores in the TRACERx 100 cohort and identified indel burden in subsets of TMB high cases. A universal mathematical law describes accuracy limitations inherent to psTMB, which result in significant misclassification rates. This scenario can be controlled by two measures: i) a panel design that is based on the mathematical law described herein: halving the CV requires a fourfold increase in panel size, ii) a novel three-tier TMB classification scheme. Moreover, inclusion of indel burden can complement TMB reports. This work has significant implications for panel design, TMB testing, clinical trials and patient management.

The growth of a tumor depends to a certain extent on an increase in mitotic events. Key steps during mitosis are the regulated assembly of the spindle apparatus and the separation of the sister chromatids. The microtubule-associated protein Aurora kinase A phosphorylates DLGAP5 in order to correctly segregate the chromatids. Its activity and recruitment to the spindle apparatus is regulated by TPX2. KIF11 and CKAP5 control the correct arrangement of the microtubules and prevent their degradation. In the present study, we investigated the role of these five molecules in non-small cell lung cancer (NSCLC). We analyzed the expression of the five genes in a large cohort of NSCLC patients (n=362) by quantitative real-time PCR. Each of the genes was highly overexpressed in the tumor tissues compared to corresponding normal lung tissue. The correlation of the expression of the individual genes depended on the histology. An increased expression of AURKA, DLGAP5, TPX2, KIF11 and CKAP5 was associated with poor overall survival (P=0.001-0.065). AURKA was a significant prognostic marker using multivariate analyses (P=0.006). Immunofluorescence studies demonstrated that the five mitosis-associated proteins co-localized with the spindle apparatus during cell division. Taken together, our data demonstrate that the expression of the mitosis-associated genes AURKA, DLGAP5, TPX2, KIF11 and CKAP5 is associated with the prognosis of NSCLC patients.

Reliable entity subtyping is paramount for therapy stratification in lung cancer. Morphological evaluation remains the basis for entity subtyping and directs the application of additional methods such as immunohistochemistry (IHC). The decision of whether to perform IHC for subtyping is subjective, and access to IHC is not available worldwide. Thus, the application of additional methods to support morphological entity subtyping is desirable. Therefore, the ability of convolutional neuronal networks (CNNs) to classify the most common lung cancer subtypes, pulmonary adenocarcinoma (ADC), pulmonary squamous cell carcinoma (SqCC), and small-cell lung cancer (SCLC), was evaluated. A cohort of 80 ADC, 80 SqCC, 80 SCLC, and 30 skeletal muscle specimens was assembled; slides were scanned; tumor areas were annotated; image patches were extracted; and cases were randomly assigned to a training, validation or test set. Multiple CNN architectures (VGG16, InceptionV3, and InceptionResNetV2) were trained and optimized to classify the four entities. A quality control (QC) metric was established. An optimized InceptionV3 CNN architecture yielded the highest classification accuracy and was used for the classification of the test set. Image patch and patient-based CNN classification results were 95% and 100% in the test set after the application of strict QC. Misclassified cases mainly included ADC and SqCC. The QC metric identified cases that needed further IHC for definite entity subtyping. The study highlights the potential and limitations of CNN image classification models for tumor differentiation.