Genetic and epigenetic intratumor heterogeneity impacts prognosis of lung adenocarcinoma

Abstract

Intratumor heterogeneity (ITH) of genomic alterations may impact prognosis of lung adenocarcinoma (LUAD). Here, we investigate ITH of somatic copy number alterations (SCNAs), DNA methylation, and point mutations in lung cancer driver genes in 292 tumor samples from 84 patients with LUAD. LUAD samples show substantial SCNA and methylation ITH, and clonal architecture analyses present congruent evolutionary trajectories for SCNAs and DNA methylation aberrations. Methylation ITH mapping to gene promoter areas or tumor suppressor genes is low. Moreover, ITH composed of genetic and epigenetic mechanisms altering the same cancer driver genes is shown in several tumors. To quantify ITH for valid statistical association analyses, we develope an average pairwise ITH index (APITH), which does not depend on the number of samples per tumor. Both APITH indexes for SCNAs and methylation aberrations show significant associations with poor prognosis. This study further establishes the important clinical implications of genetic and epigenetic ITH in LUAD.

Fig. 1. Summary of subjects and tumor samples.

Summary of tumor samples that were analyzed on different platforms: ultra-deep targeted sequencing of cancer driver genes (180 tumor samples from 56 subjects), genome-wide methylation (205 tumor samples from 68 subjects) or SNP array profiling (268 samples from 80 subjects). Top panel (bar plot): the total number of tumor samples from each subject. Bottom panel (heatmap): the number of tumor samples from each subject profiled by different platforms. The green color indicates samples assayed by the specific platform were selected for ITH analysis.

Fig. 2. Intratumor heterogeneity of single nucleotide variants for 35 lung driver genes.

Only nonsynonymous mutations are included for analysis. Top panel: the number of public (shared by all tumor samples in a patient) and private (not shared by all tumor samples in a patient) mutations for each driver gene. Bottom panel: summary of intratumor heterogeneity for each gene in each tumor. Thick lines separate the different tumors. Multiple lanes within thick lines indicate multiple samples from the same tumor. Red and orange indicate public and private mutations, respectively.

Fig. 3. Intratumor heterogeneity of somatic copy number alterations (SCNAs).

a Left panel: SCNA landscape of 80 subjects with multi-region sampling. Right panel: fractions of genomes disrupted by a specific type of SCNA. Colors represent different SCNA event types: blue indicates deletion, yellow indicates loss of heterozygosity, red indicates amplification, and white indicates copy number neutral. b APITH score (left) and naïve (right) calculated based on the tumor samples of subject IGC-11-1044 with seven tumor samples. For a given set of tumor samples, APITH was calculated as the average pairwise distance between any pair of tumor samples; naïve ITH index was calculated as the fraction of genome disrupted by private SCNAs that were not shared by all tumor samples. For a given number (K = 2, …, 7) of tumor samples, we numerated all combinations of K tumor samples to derive the distribution of ITH index. The naïve ITH index positively depends on the number of tumor samples while APITH does not. The center line in the box plots indicates median APITH or naïve ITH index. The box length indicates the interquartile range (IQR). The whiskers extend to the largest and smallest APITH or naïve ITH at most 1.5*IQR. c Distribution of pairwise average ITH of SCNAs for 80 subjects with average APITH score 0.184.

Fig. 4. Intratumoral heterogeneity of DNA methylation profiles.

a Unsupervised hierarchical clustering of 5000 most variable probes in CpG islands of the genome in 68 subjects. Different tumors are indicated by different colors in the column sidebar, with normal samples colored in gray. The numbers in parenthesis are the number of normal tissue samples for the ‘normal’ group, or the number of tumor samples in each patient. The beta values represent estimates of methylation levels, with 0 being unmethylated and 1 fully methylated. b Distribution of ITH of DNA methylation in different genomic contexts. TSS 1500: 200–1500 bases upstream of the transcription start sites (TSS), TSS 200: 0–200 bases upstream of the TSS. 5′UTR: Within the 5′ untranslated region, between the TSS and the ATG start site. Gene body: Between the ATG and stop codon. 3′UTR: From the stop codon to poly A tail. Island: CpG island. Shore: 0–2 kb from island. Shelf: 2–4 kb from island. North: upstream (5′) of island. South: downstream (3′) of island. c ITH of DNA methylation in oncogenes (n = 176), tumor suppressor genes (n = 223) and other genes (n = 12,837). The p-values are based on unpaired two-sided t-test of the two groups indicated by arrows. The center line in the box plots indicates median APITH. The box length indicates the interquartile range (IQR). The whiskers extend to the largest and smallest APITH at most 1.5*IQR.

Fig. 5. Intratumor heterogeneity of genomic and epigenomic alterations of 13 cancer driver genes in RTK/RAS/RAF pathway.

Shown are public and private SNVs, SCNAs, and DNA methylation alterations in 84 subjects. SNVs, SCNAs, and DNA methylation alterations are indicated by yellow/orange, red, and blue, respectively. Clonal and subclonal events are indicated by light and dark shades, respectively.

Fig. 6. Reconstruction of evolutionary trajectories from SCNA and DNA methylation profiles.

a Pairwise distance of tumor samples from the same tumor based on DNA methylation and SCNA profiles. Each dot represents a pair of tumor samples from the same subject. Significant Spearman’s correlation coefficient is shown (n = 212 tumor sample pairs). b Phylogenetic analysis of subject IGC-11-1044 based on DNA methylation and SCNA profiles, and the consensus phylogenetic tree built based on the distance incorporating both SCNA and DNA methylation profiles. Blue lines represent alterations shared by all tumor samples from the same subject. Yellow lines represent alterations shared by two or more tumor samples. Red lines represent alterations specific to one tumor sample. Green lines represent alterations specific to one normal sample. c Consensus phylogenetic trees for six tumors with at least five samples assayed for both SCNAs and DNA methylation. SNVs, deletions of tumor suppressor genes and amplifications of oncogenes are marked on the inferred phylogenetic tree.

Fig. 7. Kaplan–Meier curves of overall survival.

Kaplan–Meier estimates of overall survival in patients with high and low ITH of a SCNAs, b DNA methylation based on the top 5000 most variable CpG probes and c DNA methylation at CpG islands. d Kaplan–Meier estimates of metastasis in patients with high and low ITH of DNA methylation at CpG islands. High and low ITH groups were stratified by median APITH and colored in red and blue, respectively. The p-values were calculated using the Cox proportional-hazards model weighted by the variance of the estimated APITH. Sample size for each group is indicated in the figure.