Objectives Perineural invasion (PNI) is an important pathological feature of cervical cancer, which is often associated with poor prognosis. However, the underlying molecular mechanisms remain unknown.

Methods We performed whole-exome sequencing on 42 pre-treatment tumor samples (PNI:20; nonPNI:22) along with the matched blood samples and RNA-seq sequencing on 43 tumor samples (PNI:21, nonPNI:22). We performed an integrated bioinformatic analysis to identify at the gene, pathway, omics and tumor-microenvironment levels to explore the genetic determinants responsible for PNI in cervical cancer.

Results Among 45 known cervical cancer driver genes, we detected 34 in at least one patient in this cohort, including PIK3CA as the most frequently mutated genes (38%), and followed by KMT2C (19%). Comparing with nonPNI, PNI tumors harbor significantly more FBXW7 loss-of-function mutations (PNI:6; nonPNI:1, p=0.04) and copy-number gain of NKX2–1, PDGFRA (NKX2–1, p=0.007; PDGFRA, p=0.04). PNI tumors show significantly lower tumor mutation burden than non-PNI (p = 0.048). We identify 318 genes significantly dysregulated in PNI tumors relative to non-PNI tumors (upregulated: 118; downregulated: 200, |log₂FC|>1, FDR < 0.25), including downregulation of two tumor-suppressor genes, SOX17 and PTCH1. Interestingly, we find the deactivation of immune-related hallmark pathways in PNI tumors, including interferon_gamma_response, interferon_alpha_response and IL2_STAT5_signaling. Consistently, compared with nonPNI, there are significantly fewer CD8⁺ cells in the tumor microenvironment of PNI tumors (p=0.008).

Conclusions Loss-of-function mutations in FWXB7 and down-regulation of SOX17 and PITH1 are likely responsible for PNI in cervical cancer, and a tumor immunosuppressive environment may also be a contributing factor.