Introduction

Endocervical adenocarcinoma accounts for 20–25% of invasive cervical carcinomas, and its incidence is gradually increasing [1, 2]. As tumor etiology, i.e., human papillomavirus (HPV) infection, becomes an important factor for predicting the clinical outcome of endocervical adenocarcinoma, the International Endocervical Adenocarcinoma Criteria and Classification, a recent proposal for endocervical adenocarcinoma classification, divided endocervical adenocarcinoma into HPV-related and HPV-unrelated types [3]; HPV-unrelated types include gastric-type mucinous carcinoma (GAS), clear cell carcinoma, and mesonephric carcinoma. Of HPV-unrelated endocervical adenocarcinoma, GAS is the most common subtype, which accounts for ~10% of endocervical adenocarcinoma [3]. GAS is considered a distinct and clinically important entity as it is unrelated to HPV, has aggressive behavior and worse clinical outcomes than usual type endocervical adenocarcinoma, and exhibits chemoresistance [4,5,6,7,8]. Histologically, GAS is defined as a mucinous carcinoma with voluminous eosinophilic cytoplasm and distinct cell borders with glands resembling nonneoplastic gastric glands or pancreaticobiliary adenocarcinoma [9]. In addition, GAS typically shows widespread involvement and advanced stage at initial diagnosis [10]. Immunohistochemically, GAS shows positivity for HIK1083 and/or MUC6, which are components of pyloric-type mucin [11, 12]. In addition, aberrant expression of p53 is frequently observed in GAS with 41.3–64.3%, while p16 block positivity is rarely found despite HPV negativity [3, 13, 14]. While distinct morphology of GAS poses a possible genetic similarity to gastric or pancreaticobiliary cancers, the potential relationship has not been investigated.

To understand the pathogenesis of GAS and identify available therapeutic targets, its genetic characteristics and underlying pathogenic mechanisms need to be investigated. According to TCGA data of cervical cancer, the HPV-negative group showed higher epithelial–mesenchymal transition (EMT) mRNA score and significantly lower promoter-methylation level compared to HPV-positive cancer group [15]. In addition, they proposed a unique HPV-negative subgroup, which might include a considerable number of GAS cases, harboring frequent KRAS, ARID1A, and PTEN mutations similar to endometrial cancer. Recently, few studies reported targeted sequencing results of GAS. Murali et al. identified that TP53, GNAS, and SMAD4 were recurrently mutated in eight GAS cases; such mutations are also frequently identified in pancreaticobiliary or gastrointestinal cancers [16]. Garg et al. analyzed genetic alterations of 14 GAS cases and reported that known genetic variants of endometrial cancers, such as POLE, ARID1A, KRAS, and FBXW7, were enriched in GAS [13]. These findings suggest that the genetic characteristics of GAS are distinct from those of HPV-positive cervical cancers.

In this study, we performed targeted deep sequencing in 21 GAS cases to identify their genetic characteristics. Immunohistochemistry of p16, p53, and E-cadherin was performed to support the diagnosis of GAS or identify EMT-related alteration. We also compared their genetic features to those of cervical squamous cell carcinoma, endocervical adenocarcinoma, endometrial carcinoma, and gastric and pancreaticobiliary adenocarcinoma in an effort to increase the understanding of GAS pathogenesis.

Materials and methods

Patients

A total of 138 patients were diagnosed with endocervical adenocarcinoma at Yonsei Severance Hospital between June 2015 and April 2019. The cases of endocervical adenocarcinoma included 99 usual type (71.7%), 23 GAS (16.7%), 8 intestinal type mucinous carcinoma (5.8%), 3 clear cell carcinoma (2.2%), 2 mesonephric carcinoma (1.4%), 1 signet-ring cell type mucinous carcinoma (0.7%), 1 villoglandular carcinoma (0.7%), and 1 invasive stratified mucin-producing carcinoma (iSMILE) (0.7%). Among 23 patients with GAS, 21 patients with available formalin-fixed paraffin-embedded (FFPE) tissue and informed consent were investigated. Twenty-one tumor samples and 17 matched available normal samples were analyzed. Clinicopathologic data were obtained from medical records and pathology reports. This study was approved by the Institutional Review Board of Severance Hospital (IRB no. 4-2019-0538).

Histologic analysis

Hematoxylin and eosin-stained slides were reviewed by two gynecologic pathologists (EP and NHC) based on the 2014 WHO classification [17]. All of the enrolled cases showed at least focal areas with typical histology of GAS: (1) voluminous, clear, or pale eosinophilic cytoplasm; (2) distinct cell borders; and (3) moderate to marked nuclear atypia.

Immunohistochemistry

Immunohistochemistry was used to assess p16, p53, and E-cadherin expressions. Four-micrometer FFPE slices were deparaffinized and rehydrated with xylene and alcohol solutions. Immunostaining was performed using an automatic immunostaining instrument (Ventana Benchmark XT; Ventana Medical Systems, Tucson, AZ, USA) according to the manufacturer’s recommendations. The p16 (prediluted, clone E6H4; Ventana), p53 (1:300, clone DO-7; Novocastra, Leica Biosystems, Newcastle Upon Tyne, UK), and E-cadherin (1:100, clone NCH-38; Dako, Carpinteria, CA, USA) were used as primary antibodies. The p16 immunostaining pattern was interpreted as block positivity when strong, nuclear, or nuclear plus cytoplasmic staining of horizontal continuous cells was identified. For p53, at least 90% intense nuclear staining was defined as an overexpression pattern, no staining was defined as a complete absence pattern, and focal weak positive staining was defined as wild-type pattern [18, 19]. The loss of membranous expression was evaluated for E-cadherin.

HPV genotyping

HPV genotyping was performed with polymerase chain reaction (PCR)-based microarray analysis using an HPV 9G DNA chip (Biometrix Technology Inc., Chuncheon, South Korea), which detects 14 high-risk (types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68) and 5 low-risk (types 6, 11, 34, 40, and 42) HPV types. The test was performed according to the manufacturer’s protocol [20]. Briefly, genomic HPV DNA was extracted and amplified via PCR. The PCR product was mixed with hybridization solution and loaded on the HPV 9G DNA chip. DNA chip was incubated at 23–26 °C for 30 min and washed at room temperature (25 °C). Array images of the final product were automatically scanned and analyzed using a fluorescent scanner (ScanArray GX Microarray Scanner; PerkinElmer Life and Analytical Sciences, Waltham, MA, USA).

Next-generation sequencing (NGS) analysis

For DNA extraction, ten 5 µm-thick sections from each FFPE tissue were manually microdissected or laser microdissected (LMD6500, Leica Microsystems). Genomic DNA was then purified and isolated from paired tumor and normal tissue samples using the GeneRead DNA FFPE Kit (Qiagen) according to the manufacturer’s instructions.

A 468.654-kbp custom capture probe was created with the SureSelect DNA Advanced Design Wizard (Agilent Technologies, Santa Clara, CA, USA) including exons of 96 genes. The panel was based on frequently altered genes in cervical carcinoma, endometrial carcinoma, gastric, and pancreaticobiliary adenocarcinoma, recently reported mutated genes in GAS, and EMT-related genes [13, 15, 16, 21,22,23,24,25,26]. The gene list of the custom-designed panel is shown in Supplementary Table 1.

Target enrichment was performed using the SureSelect Target Enrichment protocol for the Illumina paired-end sequencing library (Agilent Technologies, ver. B.3, San Diego, CA, USA). Briefly, 200 ng of input DNA was extracted from FFPE tissue. DNA quantification and quality were measured using the PicoGreen dsDNA quantitation assay (Molecular Probes, Eugene, OR, USA). DNA was diluted in elution buffer and fragmented using Adaptive Focused Acoustic technology (AFA; Covaris, Woburn, MA, USA) according to the manufacturer’s recommendations. Fragmented DNA was repaired, adapter ligated, and PCR amplified. For exome capture, 250 ng of DNA library was mixed with hybridization buffers, blocking mixes, RNase block, and 5 µl of SureSelect all exon capture library according to the standard Agilent SureSelect Target Enrichment protocol. Hybridization to capture baits was conducted at 65 °C using a heated thermal cycler lid option at 105 °C for 24 h on the PCR machine. Captured DNA was then washed and amplified. The final purified product was quantified using quantitative PCR with the KAPA Library Quantification kit (Kapa Biosystems, Wilmington, MA, USA), and qualified using the TapeStation DNA ScreenTape D1000 (Agilent Technologies). Sequencing was performed with the HiSeq™ 2500 platform (Illumina, San Diego, CA, USA).

For data analysis, all reads were aligned to the human genome (hg19) with Burrows–Wheeler Aligner. Further data processing, variant calling, and annotation were performed according to the Genome Analysis Toolkit (GATK) best practice guidelines using Picard, GATK, and SnpSift. According to the 1000 Genomes, dbSNP, and ESP6500SI-V2 databases, variants with population frequency over 1% were grouped as SNPs and excluded. For data comparison with other cancer types, mutational data of The Cancer Genome Atlas (TCGA) dataset including 291 cervical carcinoma, 24 endocervical adenocarcinoma, 248 endometrial carcinoma, 295 gastric adenocarcinoma, 150 pancreatic ductal adenocarcinoma, and 35 cholangiocarcinoma cases were retrieved from cBioPortal.

Statistical analysis

A χ2 test or Fisher’s exact test was used to evaluate the statistical significance between two variables, and Kaplan–Meier method was used for survival analysis. Age distribution was evaluated with Shapiro–Wilk test. p < 0.05 was considered statistically significant for all analyses, and all analyses were two-tailed. All data were analyzed using R software (version 3.6.1).

Results

Patient characteristics

Among 21 patients with GAS, the median patient age was 56.6 years, and their age ranged from 32 to 90 years. FIGO stage was IB in seven (33.3%) patients, II in five (23.8%) patients, III in four (19.0%) patients, and IV in five (23.8%) patients. The median follow-up period was 16.3 months. Twenty patients (95.2%) received total hysterectomy, while one patient (4.8%) received punch biopsy for the diagnosis and was transferred for palliative care due to old age and advanced stage. At the time of initial diagnosis, p16 and p53 immunohistochemistry and HPV genotyping were performed in all cases. Seventeen (81.0%) cases were p16-negative, while four (19.0%) cases showed block positivity. The p53 staining result included 13 (61.9%) overexpression, 3 (14.3%) complete absence, and 5 (23.8%) wild-type pattern. All cases showed negative HPV genotyping results. NGS-tested specimens included 17 (81.0%) radical hysterectomy, 1 (4.8%) large loop excision of the transformation zone, and 3 (14.3%) punch biopsy cases. Patient characteristics are summarized in Table 1, and detailed clinicopathologic information of patients is shown in Table 2.

Table 1 Clinicopathologic characteristics of patients with gastric-type mucinous carcinoma.
Full size table
Table 2 Clinicopathologic details of 21 patients with gastric-type mucinous carcinoma.
Full size table

Among a total of 138 patients with endocervical adenocarcinoma, 110 patients (79.7%) were classified as HPV-related subtypes—usual type endocervical adenocarcinoma, villoglandular carcinoma, intestinal type mucinous carcinoma, signet-ring cell type mucinous carcinoma, and iSMILE; while 28 patients (20.3%) were classified as HPV-unrelated subtypes—GAS, clear cell carcinoma, and mesonephric carcinoma. Compared with HPV-related endocervical adenocarcinoma, GAS showed significantly higher initial stage, higher frequency of distant metastasis, and worse overall survival (Table 3 and Fig. 1).

Table 3 Clinical characteristics of gastric-type mucinous carcinoma compared to HPV-related endocervical adenocarcinoma.
Full size table
Fig. 1: Kaplan–Meier curve showing overall survival in gastric-type mucinous carcinoma and HPV-related endocervical adenocarcinoma.
figure 1

Patients with gastric-type mucinous carcinoma showed significant worse overall survival than those with HPV-related endocervical adenocarcinoma.

Full size image

Histologic characteristics

All cases harbored at least focal areas of classic GAS histology with abundantly clear to pale eosinophilic cytoplasm and distinct cell borders. Among them, 14.3% (3/21) of cases were entirely composed of well-differentiated tubular glands which could be designated to minimal deviation adenocarcinoma. Well-differentiated GAS resembled nonneoplastic gastric glands (Fig. 2a–c). GAS was also similar to gastric tubular adenocarcinoma and pancreatic ductal adenocarcinoma (Fig. 2d–f). Small tubular, angulated, cystic, or cribriform glands and single infiltrating tumor cells were identified (Fig. 3). Cytologically, tumor cells had voluminous, clear, foamy, or pale eosinophilic cytoplasm with mild to severe nuclear atypia. Goblet cells and abortive mucin production were not infrequently identified. In addition, GAS showed diffuse infiltration to deep cervical stroma and/or wide involvement to ectocervix and endometrium. At the invasive front of the tumor, single infiltrating tumor cells were frequently identified, and the single dyscohesive cells showed decreased expression of E-cadherin (Supplementary Fig. 1).

Fig. 2: Typical histology of gastric-type mucinous carcinoma and its histologic comparison with gastric and pancreatic lesions.
figure 2

Well-differentiated GAS resembles nonneoplastic gastric glands (a gastric-type mucinous carcinoma, b nonneoplastic gastric glands, c a high-power view of gastric pyloric glands). Moderately differentiated gastric-type mucinous carcinoma shows histologic similarity to gastric and pancreatic ductal adenocarcinoma (d gastric-type mucinous carcinoma, e gastric tubular adenocarcinoma, f pancreatic ductal adenocarcinoma).

Full size image
Fig. 3: Various histologic features of gastric-type mucinous carcinoma.
figure 3

a Infiltrating small tubular glands with abundantly clear to pale eosinophilic cytoplasm and distinct cell borders. b Dilated glands with micropapillary infolding and goblet cells. c Cribriform pattern of tumor cells with desmoplastic stromal reaction. d Single infiltrating tumor cells with severe nuclear atypia and eosinophilic cytoplasm at the invasive front of the tumor.

Full size image

Genetic characteristics

Overall, 55 nonsynonymous somatic mutations, including 41 missense, 1 nonsense, 4 inframe indel, 8 frameshift, and 1 splice site mutations, were detected in 21 GAS samples. The average mutation rate was 2.6 mutations per lesion (range: 0–9). VAF ranged from 5.8 to 75.3%, with an average read depth of 1285.60×. The most frequently mutated gene was TP53 (11/21, 52.4%), followed by serine-threonine kinase 11 (STK11/LKB1), HLA-B, PTPRS (4/21, 19.0%), FGFR4 (3/21, 14.3%), GNAS, BRCA2, ELF3, ERBB3, KMT2D, SLX4 (2/21, 9.5%), CDH1, EPCAM, KRAS, MLH1, RNF43, SNAI1, TWIST1, ZEB1, ZEB2, etc. (1/21, 4.8%) (Fig. 4). According to functional annotation, major mutant genes were categorized as involved in signal transduction, DNA damage repair, and EMT. In EMT-related genes, the mutational frequency of GAS was compared to that of cervical squamous cell carcinoma (n = 291) and endocervical adenocarcinoma (n = 24) TCGA datasets. Of eight EMT-related genes (CDH1, ELF3, EPCAM, SNAI1, SNAI2, TWIST1, ZEB1, and ZEB2), the total mutational frequency was 33.4% in GAS, while it was 9.8% and 8.4% in cervical carcinoma and endocervical adenocarcinoma, respectively (Supplementary Table 2).

Fig. 4: Genetic alterations of gastric-type mucinous carcinoma.
figure 4

Clinical data, p53 immunohistochemistry, and genetic alterations in 21 patients with gastric-type mucinous carcinoma. EMT epithelial–mesenchymal transition.

Full size image

Of TP53 mutations, 90.9% (10/11) was located in DNA-binding domain (residues 94–312) and 9.1% (1/11) was located in carboxy (C)-terminal domain (residues 355–393). With respect to the correlation between p53 immunohistochemistry and TP53 mutation status, 68.8% (11/16) of cases with abnormal p53 expression harbored correlated TP53 missense or nonsense mutations or pathogenic splice site mutation. The remaining 31.3% of cases (5/16) showed p53 overexpression, however, TP53 mutation was not identified (Supplementary Table 3 and Supplementary Fig. 2).

We compared the genetic features of GAS with frequently mutated genes of cervical squamous cell carcinoma, endocervical adenocarcinoma, endometrial carcinoma, and gastric and pancreaticobiliary adenocarcinoma based on TCGA data. The results are summarized in Fig. 5, Supplementary Fig. 3, and Supplementary Table 4–9.

Fig. 5: Mutational frequency of gastric-type mucinous carcinoma in frequently altered genes of cervical squamous cell carcinoma, endocervical adenocarcinoma, endometrial carcinoma, gastric adenocarcinoma, pancreatic ductal adenocarcinoma, and biliary adenocarcinoma.
figure 5

Representative frequently mutated genes in cervical squamous cell carcinoma, endocervical adenocarcinoma, endometrial carcinoma, gastric adenocarcinoma, pancreatic ductal adenocarcinoma, and biliary adenocarcinoma were listed in the each bar graph. Their mutational frequencies were compared with those of gastric-type mucinous carcinoma.

Full size image

Discussions

In this study, we investigated genetic characteristics of 21 GAS cases to understand the biological basis of GAS. TP53 was the most frequently mutated gene, and mutated genes were mostly involved in signal transduction, DNA damage repair, and EMT. In addition, GAS shared some genetic features with gastric and pancreaticobiliary adenocarcinoma.

Histologically, pyloric gland mucin is a characteristic feature of GAS tumor cells. Although it is normally secreted from gastric pyloric glands and mucous neck cells, pyloric metaplasia could be identified in other organs as a protective process toward chronic injury or inflammation [27]. In addition, pyloric metaplasia could have a sensitizing effect toward oncogenic stimuli, which increases the risk of malignancy [28]. In uterine cervix, the possible link between benign pyloric metaplastic lesions and GAS had been suggested [29]. If GAS develops from benign metaplastic lesions, pyloric gland phenotype of GAS might be retained from the benign precursors. On the other hand, pyloric differentiation could be a manifestation of stem cell-like properties of GAS. Cancer stem cells could transdifferentiate to other cell types to support the tumor’s aggressive behavior [30]. In our GAS cases, genes associated with stem cell properties, such as CDH1, EPCAM, SNAI1, TWIST1, ZEB1, and ZEB2, were frequently altered. Although the underlying mechanism remains unknown, the pyloric phenotype of GAS may result from complex genetic or epigenetic mechanisms and further study is needed.

In our analysis, the majority of mutated genes were involved in signal transduction and DNA damage repair, which were commonly identified mutations in most malignant tumors. In addition, a considerable number of EMT-related genes were mutated in GAS, which was higher than those of the cervical squamous cell carcinoma and endocervical adenocarcinoma TCGA dataset. Alterations of EMT-related genes are frequently identified in multiple cancers and regarded to contribute to EMT phenomenon [31]. EMT is crucial for tumor dissemination and chemoresistance [32,33,34]. In this regard, EMT-related pathways might have a role in aggressive behavior and chemoresistance of GAS [8]. TCGA mRNA data of the HPV-negative group also showed higher EMT mRNA score compared to the HPV-positive group, in line with our results [15].

In our GAS cases, 76.2% showed p53 protein overexpression; however, 31.3% (5/16) of these cases harbored wild-type TP53. Likewise, Garg et al. reported three GAS cases with p53 overexpression and wild-type TP53 [13]. The accumulation of wild-type p53 had been identified in various tumors and regarded as a response to a variety of cellular stresses [35,36,37,38,39]. DNA damaging agents such as ultraviolet light radiation and chemotherapeutic agents could induce accumulation of wild-type p53 through posttranslational modifications. In addition, oncogenic stimuli could activate and accumulate p53 by inducing DNA damage or abnormal interaction with MDM2 [40]. These cellular stresses may accelerate the rate of mutation accumulation and tumor progression in malignant tumor.

In this study, we compared the genetic features of GAS to those of other anatomically or histologically related cancers. Compared to frequently mutated genes of cervical squamous cell carcinoma, 9.5% of GAS cases harbored KMT2D mutations [15]. Regarding frequently mutated genes in endocervical adenocarcinoma, KMT2D, ERBB3, and KRAS were mutated in 9.5%, 9.5%, and 4.8% of GAS cases, respectively. Of frequently mutated genes in endometrial carcinoma, TP53 was mutated in GAS (52.4%), followed by KMT2D (9.5%) and KRAS (4.8%) [26]. Other well-known endometrial carcinoma-related genes, such as PTEN, PIK3CA, ARID1A, PIK3R1, CTNNB1, FBXW7, and POLE, were not mutated in GAS. On the contrary, Garg et al. observed frequent endometrial cancer-related gene mutations (POLE, ARID1A, KRAS, and FBXW7) in GAS [13]. One possible reason for such discrepancy is the different ethnicity of the two study populations. Although Caucasians were the majority of the patient cohort of Garg et al., our 21 GAS cases consisted exclusively of Korean population. In addition, the different genetic features between the two studies might imply genetic heterogeneity of GAS.

In comparison with gastric and pancreaticobiliary adenocarcinoma, TP53 mutations were frequently identified, near 50%, in both GAS and gastrointestinal cancers. In addition, of frequently mutated genes of gastric adenocarcinoma, KMT2D, ERBB3, and RNF43 mutations were identified in 9.5%, 9.5%, and 4.8% of GAS cases, respectively [22]. In gastric adenocarcinoma, KMT2D, ERBB3, and RNF43 mutations were associated with poor prognosis [41]. These overlapping genetic features might be a clue to  understand the distinct behavior of GAS. On the other hand, highly mutated genes of pancreatic ductal adenocarcinoma such as KRAS, SMAD4, and CDKN2A were rarely mutated in GAS [21]. These genetic differences suggest that GAS and pancreatic ductal adenocarcinoma are genetically distinct entities, despite their histologic similarities. Of frequently mutated genes in cholangiocarcinoma, BAP1, HLA-B, and TP53 mutations were identified in GAS with frequencies of 4.8%, 19.0%, and 52.4%, respectively [25]. Of these, BAP1 mutations were reported to be associated with better prognosis; however, it was correlated to small duct type cholangiocarcinoma, a subtype that is histologically dissimilar to GAS, that it was difficult to attach much meaning [42, 43].

In conclusion, this study identified a considerable number of EMT-related gene mutations in GAS, which suggest a possible role of EMT-related pathways in tumor dissemination and chemoresistance of GAS. Compared to gastric and pancreaticobiliary adenocarcinoma, TP53 mutations were frequent in both GAS and gastrointestinal cancers; KMT2D, ERBB3, and RNF43 mutations were shared between GAS and gastric adenocarcinoma, while highly mutated genes in pancreatic ductal adenocarcinoma were rarely mutated in GAS. These findings would provide clues to the biological basis of GAS.