Molecular correlates of invasion pattern in HPV-associated endocervical adenocarcinoma: emergence of two distinct risk-stratified tiers.

Abstract

The pattern-based (Silva) classification of invasive human papilloma virus (HPV)-associated endocervical adenocarcinomas (HPVA) is an established and reproducible method to predict outcomes for this otherwise stage-dependent group of tumours. Previous studies utilising targeted sequencing have shown a correlation between mutational profiles and an invasive pattern. However, such correlation has not been explored using comprehensive molecular testing. Clinicopathologic data including invasive pattern (Silva groups A, B, and C) was collected for a cohort of invasive HPVA, which previously underwent massive parallel sequencing using a panel covering 447 genes. Pathogenic alterations, molecular signatures, tumour mutational burden (TMB), and copy number alterations (CNA) were correlated with pattern of invasion. Forty five HPVA (11 pattern A, 17 pattern B, and 17 pattern C tumours) were included. Patients with pattern A presented at stage I with no involved lymph nodes or evidence of recurrence (in those with >2 months of follow-up). Patterns B and C patients also mostly presented at stage I with negative lymph nodes, but had a greater frequency of recurrence; 3/17 pattern B and 1/17 pattern C HPVAs harboured lymphovascular space invasion (LVI). An APOBEC mutational signature was detected only in Silva pattern C tumours (5/17), and pathogenic PIK3CA changes were detected only in destructively invasive HPVA (patterns B and C). When cases were grouped as low-risk (pattern A and pattern B without LVI) and high-risk (pattern B with LVI and pattern C), high-risk tumours were enriched in mutations in PIK3CA, ATRX, and ERBB2. There was a statistically significant difference in TMB between low-risk and high-risk pattern tumours (P= 0.006), as well as between Pattern C tumours with and without an APOBEC signature (P= 0.002). CNA burden increased from pattern A to C. Our findings further indicate that key molecular events in HPVA correlate with the morphologic invasive properties of the tumour and their aggressiveness. Pattern B tumours with LVI clustered with pattern C tumours, whereas pattern B tumours without LVI approached pattern A genotypically. Our study provides a biologic foundation for consolidating the Silva system into low-risk (pattern A + B without LVI) and high-risk (pattern B with LVI and pattern C) categories.