Acquisition of Genetic Aberrations During the Progression of High-Grade Intraepithelial Lesions/Micro-Invasive Squamous Cancers to Widely Invasive Cervical Squamous Cell Cancer

We searched the archive of the Diagnostic and Research Institute of Pathology, Medical University Graz, Graz, Austria, for patients with a diagnosis of HPV-induced HSIL/pT1a SCC followed by invasive cervical SCC after at least a 1-year interval. During the search period of 30 years, 8 patients with archival specimens were identified. The 1-year interval was chosen to eliminate all patients who received delayed definitive treatment. All 8 patients were either lost to follow-up after the initial diagnosis or noncompliant. There were no histologic criteria to separate de novo second tumors from direct progression. This distinction was based on resection margins in cone specimens: A new HSIL or SCC after initial resection of HSIL with clear margins was interpreted as de novo tumor, and SCC arising after biopsy or conization with involved margins was considered direct tumor progression.

All HSILs were analyzed immunohistochemically with antibodies to CK7 (Dako Omnis, FLEX, monoclonal mouse anti-human cytokeratin 7 clone OV-TL 12/30; code GA619; ready-to-use, Dako), CK17 (Dako Omnis, monoclonal mouse anti-human cytokeratin 17 clone E3, code M7046, lot 095, ready-to-use, Dako), and p63 (Dako Omnis, monoclonal mouse anti-human p63 protein, clone DAK-p63, code GA662; ready-to-use, Dako).

Next-generation sequencing libraries for mutational screening were prepared using the AmpliSeq library kit 2.0 (Thermo Fisher Scientific; cat. No. 480441) and the Ion Ampliseq Cancer Hotspot Panel V2 (Thermo Fisher, cat. no. 4475346) primer pool covering hot-spot mutations in 50 genes implicated in cancer. Sequencing was done on an Ion Genestudio S5 XL benchtop sequencer (Thermo Fisher) to a length of 200 base pairs, and initial data were analyzed using the Ion Torrent Suite Software Plug-ins (Thermo Fisher, open source, general public license, https://github.com/iontorrent/). Briefly, this included base-calling, alignment to the reference genome (HG19) using the Torrent mapper, variant-calling by a modified diBayes approach, and incorporating the flow space information. The following genes were analyzed: ABL1, AKT1, ALK, APC, ATM, BRAF, CDH1, CDKN2A, CSF1R, CTNNB1, EGFR, ERBB2, ERBB4, EZH2, FBXW7, FGFR1, FGFR2, FGFR3, FLT3, GNA11, GNAQ, GNAS, HNF1A, HRAS, IDH1, IDH2, JAK2, JAK3, KDR, KIT, KRAS, MET, MLH1, MPL, NOTCH1, NPM1, NRAS, PDGFRA, PIK3CA, PTEN, PTPN11, RB1, RET, SMAD4, SMARCB1, SMO, SRC, STK11, TP53, and VHL. Proper polymerase chain reaction amplification of all amplicons and even distribution of next-generation sequencing reads was documented in coverage analysis and plots of each sample. Called variants were annotated using open source software ANNOVAR²³  and SnpEff.²⁴  All coding, nonsynonymous mutations were further evaluated and visually inspected using the Integrated Genomics Viewer,²⁵ and variant calls resulting from technical read errors or sequence effects were excluded from the analysis.

Chromosomal copy number variations were determined by low-density whole-genome sequencing. Library preparation was performed using the NEBNext Fast DNA Library Prep Set for Ion Torrent (New England Biolabs) from 50 ng DNA according to manufacturer’s recommendations, and sequencing was done on Ion Torrent S5XL (Thermo Fisher) to a depth of approximately 5 million reads per sample. Sequence data were aligned to the reference genome (HG19), and copy number variations were called using the bioconductor package CNAnorm.²⁶  Institutional Review Board approval was obtained (31-049 ex 18/19; Medical University Graz, Graz, Austria).

We analyzed 8 women with a median age of 39 years (range, 35–57 years) at time of diagnosis of HSIL/pT1aSCC and 48 years (range, 36–75 years) at diagnosis of locally advanced SCC. Two patients with an HSIL biopsy diagnosis with no further therapy or follow-up visits (patients 1 and 4) returned with an advanced SCC 9 and 10 years later. The other 6 patients underwent cone excisions with the aim of curative therapeutic intervention. The cone excisions contained portions of the endocervical canal with a length ranging from 0.5 to 3 cm (patient 2; Figure 1, A). In 2 small flat-cone specimens an HSIL was present at the endocervical margins. The other 4 cone resections had clear margins (eg, the HSIL was completely removed). All patients had no further treatment after the cone excision before they presented with a new HSIL (patient 5) after 13 months or widely invasive SCC 4 to 24 years later (Table); regional lymph node metastases were identified in half of the patients. Symptoms at presentation were related to extensive pelvic disease with ileus and kidney failure due to ureteral stenosis, bleeding, and anemia. Half of patients (2 with and 2 without lymph node metastases) died of disease due to complications of pelvic tumor bulk 12 months (range, 9–15 months) after diagnosis of SCC; the other patients were alive after 28 months (range, 6–51 months).

All HSILs and the pT1a SCC were induced by HPV high-risk genotypes (6 × HPV16; 1 × HPV59; 1 × HPV16, HPV18, and probably carcinogenic HPV53; Table). All but 1 HSIL were solitary lesions. Patient 2 had multiple HSILs, one near the original SCJ into the endocervical canal, and a smaller focus of HSIL at the endocervical resection margin. The HSILs were separated by a long stretch of nontumorous “normal” endocervical epithelium (Figure 1, A through C). HSILs arising near the original SCJ were CK7⁻ but showed coexpression of CK17 and p63 (Figure 2, A through D; patient 3). The HSIL at the endocervical resection margin strongly expressed CK7 in the cytoplasm and p63 in nuclei, but only weak cytoplasmic CK17 staining (patient 2; Figure 3, A through D). The tall columnar epithelium toward the endocervical resection margin featured numerous individual subcolumnar reserve cells and rows of proliferating reserve cells with coexpression of CK7, CK17, and p63 (Figure 3, E through H). All HSILs were devoid of somatic mutations, but the pT1a SCC featured a PIK3CA E545K gene mutation. Chromosomal copy number variations were identified in all HSIL foci tested and the pT1a SCC, and typically affected 2 or more chromosomes. Recurrent chromosomal copy number variations were a gain of chromosome arm 3q, followed by a loss of 3p.

All 7 patients with HSIL had subsequent widely invasive SCC and carried HPV high-risk genotypes (6 × HPV16; 1 × HPV59 and HPV73; Table). The invasive SCC of patient 7 harbored an additional HPV73 (potentially carcinogenic) genotype, and HPV18 and HPV53 (probably carcinogenic genotype) were no longer present in the SCC of patient 4. Somatic mutations at high mutational frequency in hot-spot regions of cancer genes were detected in 4 SCCs (patients 2, 4, 7, 8). These included a solitary activating mutation in the PIK3CA gene in the hot-spot regions E545K (n = 2), and concomitant mutations in PIK3CA (E542K) and SMAD4 genes (n = 1) and in the HRAS and RB1 genes together with an EGFR amplification (n = 1). A solitary subclonal E545K PIK3CA gene mutation (5%) was identified in the SCC of patient 6. Patient 5, with the initial diagnosis of a PIK3CA E545K–mutated HPV16-induced pT1a SCC (resected with clear margins) returned 13 months later with a new HPV16-induced HSIL that was devoid of somatic mutations. Another 24 months later she presented with an invasive nonmetastasized SCC featuring a subclonal PIK3CA H1047R mutation. Chromosomal copy number variations were dramatically increased in all SCCs.

Four of 8 patients (patients 2, 4, 7, 8) had metastases to regional lymph nodes, and with matching driver mutations in the SCC and their corresponding lymph node metastases. Identical mutations were observed in patients 4 and 7, and the metastasis in patient 2 harbored an additional PIK3CA K111Q mutation. The metastasis of patient 8 had lost the HRAS mutation, but still featured the RB1 mutation and EGFR amplification, in addition to a newly acquired FGFR-3 mutation. The remaining 4 patients (patients 1, 3, 5, 6) had no lymph node metastases, although patients 5 and 6 featured subclonal PIK3CA mutations in their SCCs (5%, 6%, and 9% mutational frequency). Patient 1, with an initially low-stage SCC with extensive perineural involvement and LVSI, developed bilateral overexpressing p16^ink4a, HPV16⁺ ovarian metastases featuring a PIK3CA amplification. Only 1 SCC was devoid of driver mutations and metastases.

This is the first study documenting genetic aberrations in individual patients as they progressed from HSIL or pT1a SCC to widely invasive SCC during a time period of up to 24 years. For development of HPV-induced precancerous lesions, driver gene mutations were irrelevant; HPV oncogenes and chromosomal copy number variations sufficed for development of HSIL. Driver gene mutations were restricted to invasive disease, but they were present already in early invasive low-stage SCC. All activating oncogenic mutations occurred in genes governing cell survival, mainly in the PIK3CA gene.²⁷  The majority of SCCs had a solitary driver gene mutation; 2 driver mutations were the maximum. In this small cohort, driver mutations with a high mutational frequency correlated strongly with regional lymph node metastases, similar to observations in a previous study.⁷  Interestingly, the 2 SCCs with driver mutations at low mutational frequency had not metastasized to regional lymph nodes. Identical mutations in primary SCC and metastases, and in early- and advanced-stage SCC of the same patient indicate clonal disease, although 2 patients had acquired additional mutations in the metastatic clone. Although clinical treatment and prognosis of pT1a SCC without LVSI and HSIL are identical, pT1a SCCs differ on a genetic level. PIK3CA E545K–mutated tumor cells gain a fundamental biological advantage affecting cell survival and migratory behavior of tumor cells. Accumulation of genetic aberrations (eg, driver gene mutations and a gain of 3q) may predispose patients to have an even higher risk of metastases. Despite early evolution of tumor clones with driver gene mutations (eg, in microinvasive cancer), metastases apparently are late events and require a high mutational load in the primary invasive SCC. Even with driver mutations, low-stage SCC can be cured when all tumor cells are removed surgically,¹¹  but a de novo HSIL and invasive SCC with new or different mutations can develop from residual HPV-infected reserve cells and/or squamous epithelium. In this small study cohort, we were unable to confirm the postulated worse prognosis for mutated metastasized SCC.⁷  All deaths in our study were from complications of advanced-stage cancer with bulky pelvic disease, independent of mutational status or presence/absence of regional lymph node metastases. Presently, patients with SCC greater than or equal to pT1b are not candidates for primary surgical treatment, mainly because of the likelihood of metastasis. Larger studies are needed to learn if nonmutated SCCs are less likely to metastasize, if metastases represent clonal disease, and if genetic analyses are helpful in stratifying patients who may still benefit from a curative surgical approach.

The gold standard of therapy for HSIL according to present guidelines is cone excision that aims at complete removal of the HSIL, while leaving the endocervical mucosa undisturbed. When resection margins of flat-cone excisions were involved with residual HSIL, subsequent invasive SCC apparently was the result of direct progression of the HSIL.²⁸  However, even after complete resection of HSIL, women are at increased risk for SCC.²⁹  Some SCCs may arise from foci of undetected HSILs inside the endocervical canal in settings of so-called T3 transformation zones.³⁰  A de novo development of HSIL or SCC from HPV-infected residual reserve or stem cells, however, is more realistic,^12,31–33  although we were not able to determine if patients had a persistent or newly acquired HPV infection. Only one SCC featured a concomitant newly acquired HPV73 genotype that is associated with cervical cancer in rare cases.³⁴  Reserve cells can function as reservoirs of persistent HPV infections or targets of new HPV infections. Although reserve cells are distributed throughout the entire endocervical epithelium,^22,35  HSILs typically arise near the SCJ where CK17 and p63⁺ reserve cells have their highest concentration, while occasionally HSILs may also develop further up in the endocervical canal from proliferating reserve cells and immature metaplastic epithelium.¹²  When CK17/p63⁺ reserve cells are diminished or removed by cone excision for HSIL, new squamous metaplastic epithelium, the so-called neo-transformation zone, will develop from residual reserve cells, either with a CK17 or CK7 phenotype, or from both types. The dual origin of HSIL and subsequent SCC explains the inconsistent or concomitant CK17 and CK7 staining of HSIL and SCC.^36,37  With respect to cancer prevention, (residual) reserve cells should move into the center of attention. Removal of large portions of endocervical tissue (including the intraepithelial subcolumnar reserve cells) together with the HSIL achieved a near 0% recurrence rate.³⁸  This observation offers intriguing treatment concepts, such as ablation of the endocervical epithelium when treating HSILs in older women, and reserving conservative excision techniques for HSILs for women still planning a pregnancy.³⁹ 

In summary, widely invasive SCC diagnosed more than 1 year after HSIL diagnosis developed either through direct progression after incomplete treatment of HSIL, or de novo after cone excisions with clear margin via persistent or de novo HPV infection. Activating oncogenic driver mutations, predominantly in the PIK3CA gene, were restricted to invasive disease. Driver mutations at a high mutational frequency or PIK3CA gene amplifications conferred a migratory phenotype, which was evidenced by metastases, which in turn represented clonal disease. Cancer-related deaths were due to complications of bulky pelvic disease, and in this small study cohort, showed no association with driver gene mutations or metastatic disease.