Whole exome sequencing identifies clinically relevant mutational signatures in resected hepatocellular carcinoma

Most patients develop recurrent disease despite curative surgery for hepatocellular carcinoma (HCC). No standard adjuvant therapy is available and molecular predictors of outcome are poorly understood.


| INTRODUC TI ON
Hepatocellular carcinoma (HCC) is a global health problem and a leading cause of cancer-related deaths worldwide. 1,2 In particular, the poor prognosis of this disease encompasses the group of patients with localized disease at the time of diagnosis, with over half of them at risk of developing recurrent or metastatic disease despite curative liver resection and/or transplantation. Attempts to improve survival rates with adjuvant therapy following local therapy have thus far been disappointing, 3 highlighting the need to better understand the molecular aetiology of HCC with the hope of identifying novel therapeutic strategies in this era of precision oncology. 4 To date, several risk factors for HCC development have been defined, including cirrhosis, infection with hepatitis B virus (HBV) or hepatitis C virus (HCV), excessive alcohol consumption, tobacco smoking, aflatoxin exposure, as well as metabolic syndrome. 4 More recently, the genomic footprints of exposure to aristolochic acids (AA) found in plants have been discovered in liver cancers across several parts of 16

Key points
Genomic footprints of exposure to aristolochic acids (AA) found in plants have been discovered in liver cancers across several parts of Asia. In a well-annotated cohort of patients with liver cancer treated within a prospective clinical trial, signatures of AA exposure were observed in over half the cases, confirming it as a major healthcare problem. Interestingly, the identification of AA signatures was correlated with better survival outcomes, supporting their potential relevance in the clinic. Asia, and AA exposure has been shown to cause liver cancer in mice. 5,6 The AA mutational signature is characterized by induction preponderance of T > A substitutions, and have been discovered in urothelial carcinomas, renal cell carcinomas and bile duct carcinomas. 7-9 Among AA-associated HCCs, the AA signature accounts for the majority of non-silent mutations in known driver genes, and is associated with high neoantigen burden. 5 Interestingly, their implications on HCC treatment and patient outcomes remain largely unknown.
The genomic landscape of HCC has been well described, with frequent alterations in genes involved in p53 signalling, Wnt signalling, chromatin remodelling and other oncogenic pathways. [10][11][12][13][14] In particular, epidermal growth factor receptor (EGFR) has been demonstrated to be selectively overexpressed in localized HCC compared to normal hepatocytes. 15 EGFR expression was directly correlated with proliferation activity, stage, intrahepatic metastasis and histological differentiation. 15 Interestingly, the potential therapeutic relevance of targeting EGFR has been demonstrated in vitro, with EGFR tyrosine kinase inhibition gefitinib induced growth inhibition, apoptosis and cell cycle arrest. 16 In addition, in an in vivo orthotopic mouse HCC model, gefitinib has been shown to reduce tumour size and intrahepatic metastases. 17 Given these findings, we hypothesized that inhibition of EGFR through the orally administered tyrosine kinase inhibitor gefitinib may confer clinical benefit in the adjuvant setting. The initial results of our pilot clinical trial demonstrated a respectable median recurrence-free survival (RFS) of 24 months, which compares favourably with historical data. 18 In this report, we examined the genomic landscape and mutational signatures of this well-annotated cohort of patients with localized HCC, and explored their relationship with patient outcomes.

| Study design and participants
The trial was a multicentre pilot study conducted across tertiary hospitals in Singapore. Eligible patients included men and women aged 18 years or older with a confirmed histopathological diagnosis of HCC suitable for curative resection. All patients were of Chinese ethnicity. Staging was performed as per AJCC 7th edition. Before starting gefitinib adjuvant therapy, eligible patients must have complete resection with or without positive resection margins or residual disease <0.5 cm (microscopic disease). Intra-operative ultrasound was performed to rule out any residual disease, and a normal CT scan of the liver was performed 4-6 weeks after resection was required. An ECOG performance status of 0, 1 or 2 was required as well as adequate bone marrow, liver and renal functions as assessed by laboratory tests, including haemoglobin, bilirubin, platelet count, neutrophil count, alanine aminotransferase, aspartate aminotransferase and serum creatinine. Patients were required to have serum alpha-fetoprotein levels below 50 µg/L; levels above 50 µg/L were allowed for patients with hepatitis B or C cirrhosis. Exclusion criteria included the following: active infection, symptomatic heart failure, unstable angina, uncontrolled cardiac arrhythmia, psychiatric disorder, pregnancy and concomitant use of drugs which are potent inducers or inhibitors of CYP3A4. No history of other malignancy was allowed except for adequately treated basal cell or squamous cell skin cancer or cervical cancer in-situ. Written informed consent was obtained in accordance with the Declaration of Helsinki. The study is registered with ClinicalTrials.gov (NCT00282100) and approved by the IRBs of all participating hospitals. All the authors had access to the study data and had reviewed and approved the final manuscript.

| Procedures
Adjuvant treatment consisted of gefitinib 250 mg orally daily for 6 months starting within 4-6 weeks after surgery. Patients were evaluated at 1, 3 and 6 months while on gefitinib. For grade 3 or higher toxicity by NCI Common Toxicity Criteria, gefitinib was held until recovery for up to 14 days with concurrent management of the toxicities. Follow-up included serum alpha-fetoprotein levels and imaging studies every 3 months for the first 2 years, every 6 months for years 3-5 and yearly thereafter.

| Survival outcomes
The study end points were RFS, overall survival (OS) and correlative genetic studies. All statistical analyses were carried out on an intention-to-treat basis. OS was determined by the interval between date of study entry and the date of death from any cause; while RFS was determined by the interval between date of study entry and date of documented disease relapse by radiological assessment or date of death from any cause, whichever comes earliest. Survival was censored at the date of the last follow-up for survivors.

| Whole exome sequencing
Genomic DNA isolated from snap frozen tumour tissue with adequate tumour content, as well as with their paired normal tissue were selected for whole exome sequencing. A qualified pathologist provided the initial microscopic evaluation and assessment of tumour content. Whole exome sequencing was performed with hybrid selection using the Human All Exon kit SureSelect Target Enrichment System (Agilent) version 6 sequenced on the Illumina HiSeq X platform as paired-end 150-base pair reads. Read pairs were aligned to the human reference genome NCBI GRC Build 37 (hg19) using Burrows-Wheeler Aligner (BWA MEM, http://bio-bwa.sourc eforge.net/). 19 Optical duplicates were marked with Picard (http:// broad insti tute.github.io/picar d/) followed by base score recalibration using GATK version 4.1.4 for post-alignment data processing. 20 Somatic variants from the resulting normal and tumour BAM files were identified using Mutect2, and subsequently annotated and prioritized using VEP. 21 Tumour mutation burden was estimated based on the proportion of non-synonymous variants per coding megabase. Mutational signature assignment was performed with mSigAct v0.13. 5 Biologically significant copy-number changes were identified with GISTIC 2.0 and copy-number segmentations were processed with TitanCNA v1.17.1. 22,23 GenVisR was used to visualize the landscape of mutational alterations in this cohort. 24

| Statistical analysis
Receiver operating curve (ROC) analysis via the method of DeLong et al was used to derive the optimal cut-off values for continuous parameters as univariable predictors of OS. 25   and development of metastatic disease (n = 4) ( Figure S1). Table 1 summarizes   Figure 1). Under univariate analysis (Table 2), Child-Pugh score and T-stage were statistically significant predictors for OS; vascular invasion and T-stage were significant predictors for RFS ( Figure 2). Multivariate analysis (

| Correlation of molecular characteristics with survival outcomes
We examined if base substitution mutation classes correlated with survival outcomes. Across the entire subcohort with mutation data (Tables S1-S3) Both variables retained independent significance in multivariate models incorporating Child-Pugh score and T-staging (Tables S4 and   S5). Both high T > A and C > T mutation levels were correlated with TMB, though TMB by itself was not correlated with any survival outcomes ( Figure 5; Figure S6). In terms of mutational signatures, AA-mutagenesis associated with improved RFS (HR 0.37, 95% CI 0.16-0.85, P = .019) and we also observed a trend towards improved OS (HR 0.36, 95% CI 0.11-1.16, P = .088) ( Figure S6). In line with this, we noted that HCC without AA mutagenesis was associated with larger tumour size (P = .005) and early recurrences or death within 2 years of surgery (P = .037; Table S3). AA mutagenesis was neither correlated with the presence of HBV or HCV infection, nor with patient factors such as age, sex or presence of cirrhosis. EGFR amplification did not affect survival outcomes significantly ( Figure S6).  Interestingly, we observed that HCC with AA mutagenesis was smaller in tumour size and associated with a reduced incidence of early recurrences or death within 2 years of surgery.

| D ISCUSS I ON
Correspondingly, there was a suggestion of improved survival outcomes for AA-mutagenized HCC. In addition, we found that high levels of T > A transversions were associated with better patient survival. Interestingly, AA-exposed upper tract urothelial carcinomas were also recently reported to have improved survival outcomes. 33 A recent comprehensive report on the integrated proteogenomic characterization of primary resected HBV-related HCC (n = 159) described 35% of the cases harbouring AA signatures. Although the study did not show significant association with survival outcomes, AA signatures were correlated with favourable clinical features including the absence of tumour thrombus and earlier disease stages. 34 In addition, our results and others have consistently shown that HCCs with AA signatures harbour greater tumour mutation burden. 5,34 These tumours also demonstrate higher levels of predicted neoantigen load, significantly denser infiltrating CD8+ T cells, as well as higher expression of ICOS, OX40, PD-L1 and LAG3. 34 Collectively, these findings across different HCC cohorts indicate that AA signatures may confer favourable prognostic characteristics, and HCC with AA mutagenesis may potentially respond to checkpoint immunotherapy. 35 F I G U R E 5 Survival outcomes stratified by mutational signatures. A and B, Hepatocellular carcinoma (HCC) with high T > A mutations was associated with better overall survival (OS) and recurrence-free survival. C and D, Conversely, HCC with high C > T mutations was associated with worse OS. E and F, Both high T > A and C > T mutation levels were correlated with the global tumour mutation burden In conclusion, this study revealed the potential clinical relevance of mutational signatures in resected HCC. We recognize the limitations of our pilot study precluding definitive evidence for the clinical utility of gefitinib in the adjuvant setting after curative resection of HCC and acknowledge that our findings remain to be confirmed in randomized controlled studies. Nonetheless, patient outcomes with gefitinib as adjuvant therapy after curative resection for HCC were modest, highlighting the need for further research.