Mutational signatures reveal the dynamic interplay of risk factors and cellular processes during liver tumorigenesis

Nature communications 2017 AI 7 Explanations View Original
Original Paper (PDF)

Unable to display PDF. Download it here or view on PMC.

Plain-English Explanations
Pages 1-3
Reading the Mutational History of Liver Tumors

The Core Concept Every cancer carries a unique pattern of mutations - its mutational signature - that reflects the mix of forces that drove DNA damage over a lifetime. This study analyzes the whole genomes of 308 liver cancers (44 new European cases plus 264 published Japanese cases) to comprehensively catalog these signatures and understand their origins.

Why Liver Cancer? Liver cancer is unusual among cancers in having a diverse set of well-characterized risk factors: hepatitis B and C viruses, alcohol abuse, metabolic syndrome, and specific carcinogens like aflatoxin B1 and aristolochic acid. This diversity makes it an ideal model to study how different environmental exposures leave distinct mutational marks.

Ten Signatures Identified The study identified 10 mutational signatures (named by COSMIC nomenclature: Signatures 1, 4, 5, 6, 12, 16, 17, 22, 23, 24), plus 6 structural rearrangement signatures. Each has a characteristic pattern across the 96 possible single-base substitution types. The most prevalent signatures (1, 4, 5, 12, 16) collectively account for 97% of all mutations.

Beyond Simple Association Rather than just linking signatures to risk factors, the study reconstructs sub-clonal tumor architecture to reveal how mutational processes change over time as tumors develop, including dramatic examples like the disappearance of aflatoxin B1 signature after African immigrants moved to Europe.

TL;DR: Whole-genome sequencing of 308 liver cancers reveals 10 mutational signatures whose patterns are shaped by environmental exposures, DNA replication, transcription, and the dynamic history of tumor development.
Page 3
Whole Genome Sequencing and Signature Analysis

Sequencing Depth The 44 new European liver tumors were sequenced at an average depth of 92-fold for tumor tissue and 68-fold for matched non-tumor liver samples. This deep sequencing allows accurate identification of sub-clonal mutations present in only a fraction of tumor cells, enabling clonal architecture analysis.

Signature Extraction BayesNMF and EMu algorithms were used to deconvolve the mutational profiles into component signatures. Each mutation in each tumor was assigned a probability of originating from each signature based on the substitution type, trinucleotide context, and signature proportions in that tumor.

Sub-clonal Architecture Using variant allele frequencies at each mutation site, the researchers reconstructed the clonal evolution of each tumor - identifying which mutations were present in all tumor cells (clonal) versus a subset (sub-clonal), and inferring the temporal order of mutational events.

Clinical Correlation Signature intensities were correlated with clinical variables (smoking status, alcohol consumption, age, gender, viral infection status) using linear regression models, enabling statistical attribution of signature activity to specific risk factors.

TL;DR: Deep whole-genome sequencing combined with Bayesian signature extraction and clonal architecture reconstruction reveals the temporal order of mutational events in liver cancer development.
Pages 3-4
Risk Factors Modulate Specific Signatures

Alcohol and Signature 16 Signature 16, found almost exclusively in liver cancers, shows a unique pattern of transcription-coupled damage - mutations preferentially affecting the non-transcribed DNA strand at highly expressed genes. It was significantly associated with male gender (p = 1.5 x 10^-6), alcohol consumption (p = 2.0 x 10^-6), and accumulated 125 additional mutations per year of life in male drinkers versus only 12 per year in female non-drinkers.

Tobacco and Signature 4 Signature 4, linked to polycyclic aromatic hydrocarbons (PAH) from tobacco smoke, contributes an average of 1,397 mutations in smokers versus 1,004 in non-smokers. The excess in non-smokers suggests other environmental PAH sources (cooking, industrial exposure) also contribute. The slope of signature 4 accumulation with age was more than twice as steep in smokers.

Aflatoxin B1 and Signature 24 Signature 24, associated with aflatoxin B1 carcinogen exposure, was found in some African-origin patients but notably absent in tumors from African migrants living in Europe, suggesting that the signature reflects ongoing exposure rather than a permanent genomic scar - it vanishes when the exposure stops.

Mismatch Repair and Aristolochic Acid Signatures 6 (mismatch repair deficiency) and 22 (aristolochic acid from herbal medicine) were each found in only 5% of cases or fewer, consistent with their role as rare but specific carcinogens in specific patient subgroups.

TL;DR: Alcohol drives a unique transcription-coupled signature that accelerates with age, tobacco leaves PAH signatures, and aflatoxin B1 signatures disappear when exposure ends - all quantified precisely.
Pages 3-4
Signatures Drive Specific Driver Gene Mutations

Signature 16 and CTNNB1 CTNNB1 mutations (activating mutations in the Wnt pathway) were significantly more often attributable to mutational signature 16 than to other signatures (median probability 0.51 vs 0.35 in CTNNB1-wild-type tumors, p = 3.3 x 10^-10). This explains the well-known clinical observation that CTNNB1 mutations are more common in alcohol-related liver cancers.

Hotspot-Specific Attribution Different CTNNB1 mutation hotspots were predominantly caused by different signatures: most hotspots were driven by signature 16, but mutations at positions T41 and S45 were more often attributed to signature 12, showing that even within a single gene, different carcinogenic processes preferentially target different sites.

TP53 and HBV-Related Signatures TP53 mutations, more common in HBV-related cases, showed different signature attributions than CTNNB1 mutations, consistent with the biological distinction between HBV-driven (TP53) and alcohol-driven (CTNNB1) liver cancer subtypes that has been observed clinically.

Validation in Independent Cohort The association of signature 16 with male gender, alcohol, and CTNNB1 mutations was independently validated in a whole exome sequencing cohort of 573 tumors from TCGA and a previously published French series, confirming these findings are robust and not specific to the Japanese ICGC dataset.

TL;DR: Alcohol-associated Signature 16 specifically drives CTNNB1 mutations in liver cancer, explaining why CTNNB1 is enriched in alcohol-related cases - a direct mechanistic link from exposure to driver mutation.
Pages 4-5
Replication-Transcription Collisions Create Unusual Mutations

Unexpected Indels in Highly Expressed Genes The study found a surprising elevation in insertions and deletions (indels) in very highly expressed liver-specific genes (FPKM greater than 100), including albumin (ALB), apolipoprotein B (APOB), cytochrome P450 enzymes, and alcohol dehydrogenase. This contradicts the general trend of lower mutation rates in highly expressed genes.

Replication Slippage Mechanism The enriched indels were predominantly short deletions (2-5 bases) at polynucleotide repeats, characteristic of replication slippage errors. In ALB and APOB, 30% of mutations were indels versus only 3% in other genes, a 10-fold enrichment consistent with collision between the replication and transcription machineries.

Functional Implications These replication-transcription collision indels could disrupt the function of critical liver-specific metabolic proteins. However, their occurrence in genes expressed at extremely high levels may also reflect that these transcription-replication conflicts are generally tolerated by the cell unless they hit critical regulatory regions.

Broader Cancer Relevance This finding of replication-transcription collision-driven mutations in tissue-specific highly expressed genes likely occurs in other cancers as well, where the most highly expressed genes of each tissue type could similarly accumulate indels. This represents a new mechanism of oncogenesis that deserves investigation in other tumor types.

TL;DR: Liver-specific highly expressed genes accumulate unusual insertion/deletion mutations caused by collisions between replication and transcription machinery - a newly identified mechanism of genomic instability.
Pages 5-6
Clonal Architecture and the Temporal Order of Events

Sub-clonal Analysis By analyzing variant allele frequencies, the researchers could distinguish mutations present in all tumor cells (clonal, early events) from mutations present in only some cells (sub-clonal, later events). This revealed the order in which different mutational processes became active during tumor development.

Dynamic Signature Evolution In African migrants' tumors, aflatoxin B1 signature was identified in sub-clonal mutations (reflecting past exposure in Africa) but not in clonal driver mutations, suggesting the aflatoxin exposure occurred after tumor initiation rather than causing it. This remarkable finding shows the power of sub-clonal analysis to separate causation from association.

Chromosome Duplications as Late Events Chromosomal-scale duplications, which are common in HCC, were predominantly found in sub-clonal cell populations rather than in all tumor cells. This implies they occur late in tumor evolution and may represent rate-limiting steps that accelerate tumor progression after early driver mutations have established the clone.

Structural Rearrangements Six distinct structural rearrangement signatures were identified, some correlating with known mutational processes. HBV insertion sites were identified at 12 locations, consistent with HBV's known role in causing double-strand breaks that lead to chromosomal rearrangements near the insertion sites.

TL;DR: Clonal architecture reconstruction reveals that chromosome duplications are late events in liver cancer development, and aflatoxin signatures in migrants reflect past rather than ongoing exposure.
Pages 5-6
Implications for Prevention and Early Detection

Prevention Targets The quantitative dose-response relationship between alcohol consumption and signature 16 mutations (especially in males) provides strong biological rationale for alcohol reduction as a liver cancer prevention strategy. The vanishing of aflatoxin signature in migrants shows that removing exposure is sufficient to halt signature-driven mutagenesis.

Unexplained Signatures Signatures 12, 17, and 23 remain of unknown etiology in liver cancer. Identifying their causes could reveal new modifiable risk factors for liver cancer prevention. Future studies with detailed dietary, occupational, and medical histories could help decode these mysterious signatures.

Early Detection Potential Understanding which signatures initiate versus propagate tumors could point toward early detection strategies. If specific signatures leave detectable marks in circulating cfDNA early in the carcinogenic process, they could serve as early warning signals for liver cancer risk.

Therapeutic Implications Mismatch repair deficiency (Signature 6), though rare in liver cancer, is associated with response to immune checkpoint inhibitors in other cancer types. Identifying the small subset of HCC patients with mismatch repair deficiency could guide immunotherapy treatment decisions.

TL;DR: The mutational signature landscape points toward actionable prevention strategies (reducing alcohol and aflatoxin exposure) and flags unknown signatures as targets for future etiological research.
Citation: Open Access, 2017. Available at: PMC5670220.