Clinical pathological and molecular features of 100 patients with gastric-type cervical adenocarcinoma

Abstract

Objective

To investigate the clinicopathological and molecular features, diagnosis, and differential diagnosis of gastric-type cervical adenocarcinoma (GAS).

Methods

A retrospective analysis was conducted on 100 patients diagnosed with GAS at the National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences, from January 2017 to January 2025. Clinicopathological data, histological characteristics, and immunohistochemical expression patterns were analyzed. Targeted next-generation sequencing (NGS) was performed on 11 cases.

Results

The cohort comprised 100 GAS patients (median age 50 years). Common clinical manifestations included abnormal uterine bleeding and vaginal discharge, with a significant proportion presenting at advanced FIGO stages (II-IV). Histological features were characteristic, and immunohistochemistry, including markers like MUC6, p16, PAX8, and PAX2, was crucial for diagnosis and differential diagnosis. Molecular analysis of 11 cases revealed a distinct high-frequency somatic mutation profile, including TP53 (72.7%), KRAS (45.5%), SMAD4 (45.5%), CDKN2A (36.4%), PIK3CA (27.3%) and STK11 (18.2%). This profile showed molecular homology with pancreaticobiliary adenocarcinoma and was characterized by microsatellite stable (MSS) and low tumor mutational burden (TMB). Regarding molecular markers and prognosis, aberrant p53 expression was frequent (50%, 37/74) but showed no significant association with clinicopathological factors or survival outcomes (p > 0.05). In contrast, PD-L1 expression (CPS ≥ 1) was significantly associated with higher FIGO stage (p = 0.021) and shorter progression-free survival (PFS) (p = 0.046).

Conclusions

GAS is a highly malignant, HPV-independent cervical adenocarcinoma characterized by atypical clinical symptoms and complex histology. This study, representing a large cohort from Northern China, provides comprehensive insights into its clinicopathological and molecular landscape. We characterized its unique molecular profile and, importantly, identified PD-L1 (CPS ≥ 1) as a potential prognostic marker associated with shorter PFS. These findings contribute to improving diagnosis, understanding biological behavior, and identifying potential therapeutic targets for this aggressive subtype.

Background

Endocervical adenocarcinomas (ECAs) represent a significant proportion (20–32%) of all cervical malignancies and continue to pose a substantial public health challenge despite advancements in screening and therapeutic strategies [1, 2]. The 2020 World Health Organization (WHO) classification of female genital tumors introduced a revised framework, classifying ECAs into HPV-associated adenocarcinomas (HPVA) and HPV-independent adenocarcinomas (HPVIA) based on their pathogenesis [3]. Among HPVIA subtypes, gastric-type endocervical adenocarcinoma (GAS) is the most prevalent and clinically aggressive variant, often exhibiting resistance to conventional therapies and associated with a poorer prognosis compared to HPVA [4].

The diagnosis of GAS presents considerable challenges due to its variable histological appearance, which can resemble both benign conditions like lobular endocervical glandular hyperplasia and metastatic adenocarcinomas from other sites, particularly the gastrointestinal tract [5]. While histomorphological evaluation remains the cornerstone, immunohistochemical markers play a crucial role in confirming the diagnosis and differentiating GAS. However, the optimal panel and the precise expression patterns of various markers in large GAS cohorts, especially in specific populations, warrant further investigation.

Beyond histopathology and standard immunohistochemistry, a comprehensive understanding of the molecular landscape of GAS is essential for elucidating its aggressive behavior and identifying potential therapeutic targets. Emerging evidence suggests that GAS harbors a distinct molecular profile compared to HPVA and other gynecological malignancies [6, 7]. However, systematic studies integrating detailed molecular signatures with clinicopathological features and clinical outcomes in well-characterized GAS cohorts are still limited. The prognostic significance of specific molecular alterations and protein expression patterns, such as TP53 mutations and PD-L1 expression, remains to be fully established in GAS. While TP53 mutations are known drivers in many cancers, their specific prognostic role in GAS is unclear. PD-L1 expression is a key biomarker for immunotherapy, but its association with prognosis and its predictive value in the context of GAS's typically immune-cold microenvironment (MSS, low TMB) require further exploration.

Addressing these gaps in knowledge is critical for improving diagnostic accuracy, refining prognostic stratification, and developing targeted therapeutic strategies for GAS. Therefore, this study aimed to provide a comprehensive analysis of the clinicopathological and molecular features of a large cohort of GAS patients from Northern China, systematically investigating their associations with clinical outcomes to enhance our understanding of this aggressive disease.

Materials and methods

Patients and tissue samples

A retrospective cohort of 100 patients with newly diagnosed and recurrent GAS was assembled from pathological specimens obtained from the Cancer Hospital Chinese Academy of Medical Sciences between January 2017 and January 2025. The samples included colposcopic biopsies, cervical conization samples, simple hysterectomy samples, radical hysterectomy samples, and tumor cytoreductive surgery samples. All enrolled patients were confirmed to be HPV-negative prior to histopathologic diagnosis, consistent with the criteria for HPV-independent adenocarcinoma. HPV status was determined by p16 immunohistochemistry (negative or focal expression) and/or high-risk HPV molecular testing (PCR-based assays or co-testing with ThinPrep cytology test and HPV DNA analysis). For cases with diffuse p16 expression, molecular HPV testing was performed to confirm HPV negativity. Clinical staging was performed according to the 2018 International Federation of Gynecology and Obstetrics (FIGO) staging system for cervical cancer. The study cohort consisted of patients aged 25–73 years, with a mean age of 50 years. Histopathological confirmation was achieved through a blinded independent review conducted by two gynecological pathologists (S.G. and Y.G.), who examined hematoxylin‒eosin (HE) staining slides without prior knowledge of the clinical data. Tumor classification adhered to the diagnostic criteria outlined in the 2020 WHO Classification of Female Genital Tumors. Clinicopathological parameters, including patient age and clinical stage, were systematically retrieved from the electronic clinical information system database maintained by the Cancer Hospital, Chinese Academy of Medical Sciences.

HE staining and immunohistochemistry

All the samples were fixed in 10% neutral buffered formalin, subjected to standard dehydration protocols, and embedded in paraffin. Tissue sections, each 4 μm thick, were prepared for HE staining and immunohistochemical analysis via the Roche BenchMark ULTRA automated staining platform. The antibody panel comprised the following: ER (clone: SP1), HER2 (clone: 4B5), p16 (clone: E6H4), PR (clone: 1E2), and PD-L1 (clone: SP263) from Roche Diagnostics; CA125 (clone: OC125), CA19-9 (clone: C241:5:1:4), CK7 (clone: RN7), CDX2 (clone: EP25), CK20 (clone: EP23), CAIX (clone: Poly), Villin (clone: EP163), HNF1β (clone: OTIR2E9), PAX2 (clone: GR007), MUC6 (clone: MRD220), and HIK1083 (clone: M-GGMC-1) from Zhongshan Golden Bridge Biotechnology; Ki-67 (clone: MIB-1), PAX8 (clone: EP298), P53 (clone: DO-7), Napsin A (clone: MX015), and SATB2 (clone: OTI5H7) from Maixin Biotech; and CEA (clone: II-7), MLH1 (clone: ES05), MSH2 (clone: FE11), PMS2(clone:EP51) and MSH6 (clone: EP49) from Dako.

Positivity for ER, PR, P53, Ki-67, PAX2, and PAX8 was defined by nuclear localization. P53 expression patterns were categorized as either mutant type (≥ 80% diffuse nuclear positivity, complete negativity, or cytoplasmic-only staining) or wild type (< 80% heterogeneous nuclear staining). Membrane and cytoplasmic markers, including CA125, CA19-9, Villin, and CEA, require combined localization, whereas CK7, HIK1083, MUC6, and CK20 exhibit cytoplasmic specificity, with HIK1083 positivity defined as > 5%. HER2 scoring was conducted in accordance with the College of American Pathologists (CAP) gastric carcinoma guidelines. Mismatch repair (MMR) status was determined by the nuclear loss of ≥ 1 protein (MLH1, PMS2, MSH2, MSH6) for dMMR classification, whereas intact expression was defined as pMMR, with internal positive controls mandatory. The combined positive score (CPS) was used for the evaluation of PD-L1 expression in solid tumors. A cutoff score of ≥ 1 for CPS was used to define PD-L1 positivity.

A comprehensive analysis of clinicopathological parameters, including FIGO stage, histopathological features, tumor differentiation, depth of cervical stromal invasion, LVSI, nodal and metastatic status, and postoperative management, was performed. All HE staining and immunohistochemical slides were subjected to blinded review by two pathologists to ensure diagnostic consensus.

Hybrid capture-based targeted NGS

Formalin-fixed paraffin-embedded (FFPE) tissue blocks were selected on the basis of corresponding HE staining slides with pathological confirmation performed by certified pathologists according to the aforementioned evaluation criteria. Genomic DNA was isolated from selected FFPE tissue blocks via the QIAamp DNA FFPE Tissue Kit (Qiagen, Germany). DNA concentration and integrity were evaluated via a Qubit 3.0 fluorometer (Thermo Fisher Scientific, USA) and verified via 1% agarose gel electrophoresis. The DNA NGS workflow involved genomic DNA fragmentation, library preparation, and sequencing on the Illumina NextSeq 550 platform (Illumina, USA). Driver mutations were detected via an internally developed molecular diagnostic management system. Variant calling was performed with a minimum variant allele frequency threshold of ≥ 2% and a sequencing depth of ≥ 1000 × to ensure analytical validity.

Statistical analysis

Statistical analyses were conducted via SPSS 27.0 software. Categorical variables are expressed as frequency counts and percentages, and between-group comparisons were performed via the chi-square (χ²) test. Progression-free survival (PFS) was censored for patients who did not record the date of progression as months to new tumor event (progression) after initial diagnosis. Overall survival (OS) was calculated from the date of initial diagnosis to the date of death or last follow-up. The survival log-rank test and Kaplan–Meier survival curves were used to compare PFS and OS between different patient groups. A two-tailed p value < 0.05 was considered statistically significant.

Results

Clinicopathological features and histological analysis of GAS

The cohort consisted of 100 GAS patients with a median age of 50 years (range: 25–73 years). The initial clinical manifestations included abnormal vaginal bleeding (85%, 85/100), vaginal discharge (60%, 60/100), and pelvic pain (20%, 20/100). Additionally, 15% (15/100) of the patients were diagnosed incidentally during routine gynecological examinations because of the detection of cervical masses. Histopathological confirmation was obtained from various sample types including biopsies, conization, and hysterectomy specimens. Clinicopathological characteristics of the cohort are summarized in Table 1. FIGO staging (2018 version) was available for 89 patients, with a significant proportion presenting at advanced stages (Stage III: 28.1%, 25/89; Stage IV: 29.2%, 26/89). Notably, high-risk HPV testing was performed using either PCR-based molecular assays or co-testing with ThinPrep cytology test (TCT) and HPV DNA testing. All patients were confirmed HPV-negative. Important prognostic factors were also assessed where data were available. Among patients with evaluable data, lymph node metastasis was observed in 40.3% (25/62), parametrial invasion in 12.9% (9/70), and ovarian metastasis in 31.6% (24/76). Deep stromal invasion was present in 79.2% (57/72) of cases, and lymphovascular invasion was identified in 51.4% (38/74). Postoperative radiotherapy or chemotherapy was administered to a high proportion of patients (92.2%, 83/90). Detailed frequencies and percentages for all clinicopathological parameters, including those with varying numbers of assessed cases (NA), are presented in Table 1.

Table 1.

Descriptive statistics of the study cohort (n = 100)

Clinicopathologic parameter	n/Range	Percentage (%) /Median (age)
Age	25–73	50
HPV status
Negative	100	100
Positive	0	0
FIGO stage(2018 version)
I	13	14.6
II	25	28.1
III	25	28.1
IV	26	29.2
NA	11	-
Deep stromal invasion
Yes	57	79.2
No	15	20.8
NA	28	-
Lymphovascular invasion
Yes	38	51.4
No	36	48.6
NA	26	-
Distant metastasis
Yes	24	24.0
No	76	76.0
Lymph node metastasis (LNM)
Yes	25	40.3
No	37	59.7
NA	38	-
Ovarian metastasis
Yes	24	31.6
No	52	68.4
NA	24	-
Parametrial invasion
Yes	9	12.9
No	61	87.1
NA	30	-
Radiotherapy/chemotherapy
Yes	83	92.2
No	7	7.8
NA	10	-

GAS exhibited a heterogeneous morphology characterized by irregularly sized and shaped, branching, or cystically dilated glands. Well-differentiated areas retained a relatively intact glandular architecture, frequently displaying "sawtooth" or "antler-like" luminal contours, resembling the morphology of the gastric pyloric gland (Fig. 1A). In contrast, poorly differentiated regions exhibited glandular fusion and back-to-back arrangements (Fig. 1B), occasionally progressing into nested growth patterns (Fig. 1C) with a desmoplastic stromal response. The glandular lumina appeared distended by abundant eosinophilic or pale-staining mucin, with extracellular mucin extravasation forming stromal "mucin lakes" in select cases (Fig. 1D), necessitating differentiation from metastatic mucinous adenocarcinomas. Neoplastic cells displayed abundant eosinophilic or clear cytoplasm resembling gastric-type mucinous cells, frequently exhibiting apical snouting, which is characterized by a distinctive protrusion of apical cytoplasm with mucin secretion (Fig. 1E). While partial cellular polarity persisted in well-differentiated zones, a progressive loss of architectural organization was observed, which was correlated with decreased differentiation. Poorly differentiated foci exhibited marked nuclear atypia, including nuclear enlargement, hyperchromasia, increased nuclear‒cytoplasmic ratios, prominent nucleoli, and frequent mitotic activity (> 5 mitoses per 10 high-power fields [HPF]) (Fig. 1F).

Fig. 1.

Histopathological features of gastric adenocarcinoma of the cervix (GAS). A Well-differentiated areas exhibited relatively intact glandular structures, with lumina displaying characteristic "serrated" or "antler-like" branching, resembling pyloric glands of the stomach (× 200 magnification). B Poorly differentiated regions demonstrated glandular fusion and a back-to-back arrangement of glands (× 200 magnification). C Nested architectural patterns may be observed, often accompanied by a desmoplastic stromal reaction (× 200 magnification). D Glandular lumina contained abundant eosinophilic or lightly stained mucin. In certain cases, mucin extravasation into the stroma results in "mucin lakes" (× 100 magnification). E Tumor cells displayed abundant eosinophilic or clear cytoplasm, frequently exhibiting an "apical mucin secretion" pattern (× 200 magnification). F Poorly differentiated areas revealed marked nuclear atypia, characterized by enlarged, hyperchromatic nuclei, an increased nuclear‒cytoplasmic ratio, prominent nucleoli, and frequent mitotic figures (× 200 magnification)

Immunohistochemical analysis of GAS

Among the 100 GAS cases analyzed, key immunohistochemical findings were categorized as follows: p16 protein expression was not detected in 94.7% of the cases (71/75), whereas only 5.3% (4/75) exhibited diffuse strong positivity (Fig. 2A). All four cases exhibiting diffuse p16 immunoreactivity were confirmed negative for high-risk HPV by molecular testing performed prior to histopathologic diagnosis. Mutant p53 expression (diffuse strong or null staining) was detected in 50% of the patients (37/74) (Fig. 2B). ER expression was observed in 10.0% of the cases (5/50), and PR expression in 6.0% of the cases (3/50). All positive cases exhibited focal and weak nuclear staining, rather than diffuse strong positivity. Nuclear PAX8 expression was present in 74% of the samples (37/50), whereas PAX2 expression remained consistently negative in all the samples.

Fig. 2.

Immunohistochemical features of gastric adenocarcinoma of the cervix (GAS). A Negative immunohistochemical staining for p16, indicative of HPV-independent tumorigenesis (× 40 magnification). B Aberrant p53 expression was consistent with a missense mutation pattern (× 100 magnification). C Strong diffuse positivity for MUC6, supporting gastric differentiation (× 100 magnification). D PD-L1 expression with a high combined positive score (CPS) of 80, suggesting potential sensitivity to immunotherapy (× 100 magnification)

Intestinal differentiation markers

CK7 was strongly and diffusely positive in all patients (53/53, 100%), whereas CK20 expression was low (9/48, 18.8%). CDX2 was detected in 64.4% (29/45) of the patients, whereas Villin exhibited focal positivity in 36.4% (4/11) of the patients. CEA was strongly positive in 77.8% of the patients (42/54), whereas SATB2 was completely absent (0/23). CA19-9 displayed focal staining in a single sample (1/1), and CA125 positivity was observed in 60% (3/5) of the tested samples.

Gastric-type differentiation markers

MUC6 was strongly expressed in 98.4% of the patients (62/63, Fig. 2C). HIK1083 positivity was detected in 66.7% (4/6) of the patients, with CAIX demonstrating universal expression (6/6, 100%).

Clear cell carcinoma discriminators

HNF1β positivity was detected in 92.9% of the patients (13/14), whereas Napsin A was consistently negative (0/18).

Therapeutic biomarkers

HER2 overexpression (3 +) was detected in 6.9% of the patients (2/29). PD-L1 positivity (combined positive score [CPS] ≥ 1) was detected in 55.2% (16/29) of the patients, including one patient with a CPS = 80 (Fig. 2D), who demonstrated only a mild-to-moderate response to immunotherapy. Tumor proliferation activity exhibited significant heterogeneity, with Ki-67 indices ranging from 10 to 90% (median: 50.0%). Immunohistochemical analysis of mismatch repair proteins (MLH1, MSH2, MSH6, and PMS2) was conducted in 22 cases, with intact nuclear expression noted in all the tested samples, confirming a pMMR status.

Molecular characteristics

NGS analysis of 11 cases revealed recurrent somatic mutations in critical oncogenic pathways, including TP53 (72.7%, 8/11), KRAS (45.5%, 5/11), SMAD4 (45.5%, 5/11), CDKN2A (36.4%, 4/11), PIK3CA (27.3%, 3/11), and STK11 (18.2%, 2/11). Furthermore, these cases demonstrated a low TMB (mean 2/Mb). Notably, no HER2 amplification or microsatellite instability-high (MSI-H) status was observed within the cohort (Fig. 3). This NGS cohort did not include the two patients who showed HER2 3 + overexpression by immunohistochemistry.

Fig. 3.

Summary of the molecular features of gastric adenocarcinoma of the cervix (GAS)

Survival outcomes

The follow-up period ranged from 6 to 87 months, with a median follow-up of 46 months. During this period, 44 deaths (44/100,44%) were observed. Overall survival (OS) ranged from 6 to 85 months (mean, 43 months). Progression-free survival (PFS) ranged from 3 to 84 months (mean, 36 months).

The association of p53 mutation status with clinical stages and pathological risk factors

Among the 74 patients assessed for p53 mutational status, the following clinicopathological data were available: FIGO stage (69/74, 93.2%), depth of myometrial invasion (56/74, 75.7%), lymphovascular space invasion (LVSI) (59/74, 79.7%), and lymph node metastasis (LNM) (50/74, 67.6%). No statistically significant associations were identified between p53 mutation status and FIGO stage (p = 0.949), depth of myometrial invasion (p = 0.716), LVSI (p = 0.240), or LNM (p = 0.729) (Table 2).

Table 2.

Correlations between the p53 mutation status and clinicopathological risk factors

	p53		p value
	Mutant-type	Wild-type
High FIGO stage (III, IV) (2018 version)
Yes	21	19	0.949
No	15	14
Deep stromal invasion
Yes	24	23	0.716
No	4	5
lymphovascular space invasion (LVSI)
Yes	17	12	0.240
No	13	17
lymph node metastasis(LNM)
Yes	11	9	0.729
No	15	15

The association of programmed death-ligand expression (PD-L1) with clinical stages and pathological risk factors

Of the 29 patients with available PD-L1 status data, FIGO staging information was available for 27 (93.1%), and data on deep stromal invasion, lymphovascular space invasion, and lymph node metastasis were available for 22 (75.9%). The tumor PD-L1 expression (by CPS) was associated with higher clinical stage (p = 0 0.021); however, no significant association was found with other pathological risk factors, including depth of myometrial invasion (p = 0.571), LVSI (p = 0.149), or LNM (p = 0.670) (Table 3).

Table 3.

The association of programmed death-ligand (PD-L1) expression with clinical stages and pathological risk factors

	PD-L1		p value
	+	-
High FIGO stage (III, IV) (2018 version)
Yes	10	3	0.021
No	4	10
Deep stromal invasion
Yes	11	8	0.571
No	1	2
lymphovascular space invasion (LVSI)
Yes	10	16	0.149
No	1	5
lymph node metastasis(LNM)
Yes	7	4	0.670
No	5	6

Programmed death-ligand expression (PD-L1) negatively correlated with progression-free survival in gastric-type cervical adenocarcinoma, whereas p53 mutant status showed no association with survival

Seventy-four patients with p53 testing and 29 patients with PD-L1 testing had complete follow-up data. Kaplan–Meier survival analysis revealed that patients with PD-L1-positive tumors (by CPS) had significantly shorter PFS compared to those with PD-L1-negative tumors (p = 0.046) (Fig. 4). While OS was lower in patients with PD-L1-positive tumors, this difference did not reach statistical significance (p = 0.282) (Fig. 4). No significant association with PFS or OS was found in p53 mutant GAS (p = 0.342 and p = 0.505, respectively) (Fig. 4).

Fig. 4.

Kaplan–Meier curves of progression-free survival (PFS) and overall survival (OS) in patients with gastric-type endocervical adenocarcinoma. A, B PFS and OS in programmed death-ligand 1 (PD-L1)-positive and -negative groups by combined positive score (CPS). C, D PFS and OS in p53 mutant status

Discussion

The 2020 WHO classification categorizes endocervical adenocarcinomas into HPV-associated and HPV-independent subtypes, with gastric-type adenocarcinoma (GAS) being the most prevalent and aggressive HPVIA variant [3].

In our cohort of 100 GAS patients, the median age was 50 years, and common symptoms included abnormal vaginal bleeding and discharge. A significant proportion presented at advanced stages (III-IV), consistent with the highly invasive nature of GAS and challenges in early diagnosis [4, 8]. Histologically, GAS is characterized by irregular glands lined by cells with abundant eosinophilic/clear cytoplasm and apical mucin secretion, often forming stromal mucin lakes. While well-differentiated areas can mimic benign lesions, poorly differentiated areas show marked architectural and cytological atypia. Diagnosis relies on characteristic morphology, supported by immunohistochemistry [9]. Previous studies have demonstrated that the grading of GAS lacks prognostic value [10, 11].

Immunohistochemistry is crucial for the diagnosis and differential diagnosis of GAS. Consistent with the HPV-independent nature of GAS, p16 expression was largely absent (94.7%), with rare diffuse positivity observed in HPV-negative cases, highlighting its utility in distinguishing HPVIA from HPVA [12–14]. Besides, diffuse p16 expression in cervical adenocarcinoma does not necessarily indicate the presence of oncogenic HPV; it may result from an unknown mechanism causing abnormalities in the Rb pathway [13].

ER and PR expression were infrequent and weak (10.0% and 6.0%, respectively), supporting differentiation from endometrial adenocarcinoma which typically shows strong diffuse positivity [14].

The TP53 gene is a crucial tumor suppressor gene, and its mutations or aberrant expression are common in various tumors, often associated with more aggressive biological behavior and poorer prognosis [15]. In GAS, TP53 mutation rates are relatively high and it is considered a significant driver gene. Our study observed aberrant p53 expression in 50% (37/74) of patients, a frequency consistent with previous studies [16]. However, unlike TP53 mutations often associated with advanced or aggressive features in other tumor types, our study found no statistically significant association between p53 status (mutant-type vs. wild-type) and clinicopathological risk factors such as FIGO stage, deep stromal invasion, lymphovascular space invasion (LVSI), or lymph node metastasis (LNM) (Table 2, p > 0.05 for all comparisons).

Furthermore, survival analysis including 74 GAS patients revealed no statistically significant association between p53 status and either OS or PFS ( p = 0.505 and p = 0.342, respectively). This finding aligns with the observed lack of association between p53 status and clinicopathological features.

Regarding the role of p53 in GAS prognosis, previous studies are relatively limited and show inconsistent results. One study from Southern China, which included 58 GAS patients, also did not find a significant association between TP53 mutation and survival outcomes [16]. Our study, representing a larger cohort from Northern China, shows consistent results with that study, potentially suggesting that in the Chinese population, the prognostic significance of TP53 mutation in GAS may be less prominent compared to its role in tumor development. A meta-analysis of all cervical cancer subtypes found that aberrant p53 expression was associated with poorer prognosis in Caucasian patients but not in Asian populations. Our GAS-specific findings partially support this observation across different ethnicities and regions.

This finding may be attributed to several factors. Firstly, GAS itself is a highly aggressive tumor subtype, and its prognosis is primarily influenced by strong clinicopathological factors such as FIGO stage, lymph node metastasis, and treatment modality [8, 10, 17]. The impact of these factors might overshadow the role of p53 as an independent prognostic marker. Secondly, our survival analysis cohort has a relatively limited sample size (n = 74), which may result in insufficient statistical power to detect a significant association.

Although our study did not find a significant association between p53 status and survival outcomes in GAS patients, this does not necessarily imply that p53 is unimportant in GAS. Aberrant p53 may indirectly influence prognosis by affecting other signaling pathways or being associated with sensitivity to specific treatments. Future studies with larger cohorts are needed to further validate the prognostic value of p53 in GAS and to explore whether it can predict response to specific therapies.

Beyond p53, we explored the clinical and prognostic significance of PD-L1 expression in GAS. PD-L1 is a key immune checkpoint molecule and a predictive biomarker for immune checkpoint inhibitor efficacy in various cancers, including cervical cancer [18]. Prior studies suggest that positive PD-L1 expression is often linked to more aggressive behavior and poorer prognosis [19–22].

In our GAS cohort, positive PD-L1 expression (CPS ≥ 1) was significantly associated with higher FIGO stage (p = 0.021) and shorter PFS (p = 0.046). This suggests PD-L1 expression may serve as a marker for tumor aggressiveness and a potential prognostic factor in GAS.

Previous studies on the prognostic role of PD-L1 in GAS are limited. One study reported an association between PD-L1 expression (assessed by Tumor Proportion Score, TPS) and PFS, but found no significant correlation using CPS [16]. In contrast, our study, employing the CPS scoring system widely used and recommended for PD-L1 assessment in cervical cancer, demonstrated a significant association between PD-L1 (CPS ≥ 1) and PFS in our cohort. This finding provides novel evidence for the utility of PD-L1 (CPS) in GAS prognostic assessment and suggests that the prognostic significance of PD-L1 in GAS may be better captured by the comprehensive expression in the tumor microenvironment rather than tumor cell expression alone.

However, despite the association of PD-L1 positivity with poorer prognosis, its value as a predictive marker for immunotherapy efficacy in GAS warrants cautious evaluation. Our molecular analysis revealed that all tested cases were microsatellite stable (MSS) with low tumor mutational burden (TMB). Prior studies indicate that MSS and low TMB tumors are often associated with an 'immune-cold' phenotype and limited response to immune checkpoint inhibitor monotherapy [23–25]. Coupled with our observation of limited response in a patient with high CPS, PD-L1 positivity in GAS may primarily reflect intrinsic tumor aggressiveness rather than a highly immune-activated microenvironment. Thus, the predictive value of PD-L1 in GAS requires further validation, and combination therapeutic strategies may need to be explored.

Markers of Müllerian origin, such as PAX8, were frequently positive (74%), aiding in distinguishing GAS from gastrointestinal adenocarcinomas and primary ovarian mucinous adenocarcinoma, which are typically PAX8-negative [26, 27]. Conversely, PAX2 was consistently negative, supporting its utility in differentiating GAS from benign gastric-type adenomatous growth [28–30].

We assessed a panel of markers for gastric and intestinal differentiation. Gastric markers, including MUC6 (98.4% positive) and CAIX (100% positive), were highly expressed, confirming gastric differentiation [5, 9, 31–33]. While CK7 was universally positive, intestinal markers such as CK20 (18.8%), CDX2 (64.4%), and Villin (36.4%) showed variable positivity, and SATB2 was consistently negative. As reported previously, the overlapping expression patterns of these markers with pancreaticobiliary adenocarcinomas limit their utility in this specific differential diagnosis [14].

For differentiation from clear cell carcinoma, HNF1β was frequently positive (92.9%), consistent with its lack of specificity [14, 33, 34]. However, Napsin A was consistently negative, supporting its value in distinguishing GAS from clear cell carcinoma.

We assessed HER2 expression, identifying overexpression (3 +) in 6.9% (2/29) of tested patients. While HER2 is a therapeutic target in other adenocarcinomas like gastric cancer, its prevalence in GAS appears lower [35]. Notably, TP53 mutations were absent in our HER2-positive cases, potentially suggesting distinct oncogenic pathways.

To gain deeper insights into the molecular characteristics of GAS, we performed targeted next-generation sequencing (NGS) analysis on 11 patients. While the sample size is relatively limited (n = 11), these molecular data offer valuable insights given the rarity of GAS. The analysis revealed a high frequency of somatic mutations primarily affecting genes such as TP53 (72.7%), KRAS (45.5%), SMAD4 (45.5%), CDKN2A (36.4%), PIK3CA (27.3%), and STK11 (18.2%), which are involved in key oncogenic signaling pathways.

In comparison to other female genital tract tumors, our GAS molecular profile exhibited distinct differences from HPV-associated cervical adenocarcinoma and endometrial cancer. For instance, TP53 and KRAS mutation frequencies were significantly higher in our cohort, and common endometrial cancer alterations like PTEN and POLE mutations were not observed. Furthermore, the high-frequency mutated genes observed (particularly KRAS, TP53, SMAD4, and CDKN2A) demonstrated significant molecular homology with the core driver genes of pancreaticobiliary adenocarcinomas. The presence of STK11 mutations also aligns with characteristics of tumors associated with Peutz–Jeghers syndrome [36]. These unique molecular features likely contribute to the aggressive biological behavior of GAS and highlight its distinct pathogenesis compared to other cervical adenocarcinoma subtypes.

Regarding the discrepancy between HER2 expression and NGS results, we detected HER2 overexpression (3 +) by IHC in 6.9% (2/29) of tested patients, while no ERBB2 (HER2) gene amplification was detected in the 11 patients who underwent NGS analysis. It is important to clarify that the patient groups who underwent IHC and NGS analyses were not entirely overlapping. Specifically, the two patients identified with HER2 3 + overexpression by IHC were not included in the subset of 11 patients who underwent NGS analysis.

The absence of HER2 amplification and microsatellite instability-high (MSI-H) status in the NGS cohort, combined with the previously discussed MSS and low TMB characteristics, further supports the notion that GAS may have limited sensitivity to immune checkpoint inhibitor monotherapy. These molecular findings contribute to understanding the aggressiveness of GAS and suggest potential therapeutic targets, although their immunomicroenvironment characteristics may limit response to certain therapies.

This study has several limitations that should be considered. Firstly, its retrospective and single-center nature inherently carries risks of selection bias and limits the generalizability of our findings to other populations or institutions. Incomplete data for certain clinicopathological parameters also affected the completeness of baseline characteristics and the sample size available for correlation analyses with molecular markers. Furthermore, the number of cases tested for different immunohistochemical markers varied. This heterogeneity in sample size stemmed from the retrospective design, where marker testing was performed based on clinical diagnostic needs rather than a uniform research protocol. This approach may introduce selection bias and affect the reliability of the reported positivity rates. Regarding molecular analyses, the NGS results are based on a limited sample size. While providing preliminary insights into the molecular features of this rare tumor, the generalizability of these genomic findings is restricted. Similarly, the sample sizes for correlation analyses between molecular markers and clinicopathological features or survival outcomes were relatively small, potentially limiting the statistical power to detect significant associations.

Conclusion

We characterized the unique clinicopathological features and immunohistochemical profile of GAS, which are crucial for accurate diagnosis and differential diagnosis. Molecular analysis further revealed a distinct high-frequency somatic mutation profile in GAS, showing significant differences from HPV-associated cervical adenocarcinoma and molecular homology with pancreaticobiliary adenocarcinoma, potentially explaining the aggressiveness of GAS. We also observed that GAS exhibits characteristics of MSS and low TMB. Regarding the prognostic significance of molecular markers, our study found that TP53 mutations are frequent in GAS but showed no statistically significant association with clinicopathological features or survival outcomes. However, we found that PD-L1 expression (CPS ≥ 1) was significantly associated with higher FIGO stage and PFS, suggesting that PD-L1 expression is a marker reflecting tumor aggressiveness and has potential prognostic value. Overall, as a large-scale GAS cohort study from Northern China, this study systematically integrated multi-dimensional features, revealing a unique molecular profile and, using the clinically relevant CPS score, demonstrating a significant association between PD-L1 expression and shorter PFS in GAS patients. These findings provide novel insights into the diagnosis, prognostic assessment, and potential therapeutic targets for this aggressive subtype. Future studies with larger sample sizes are needed to validate these findings and further explore individualized treatment strategies based on molecular features.

Abbreviations

CA125

Cancer antigen 125

CA19-9

Carbohydrate antigen 19–9

CAIX

Carbonic anhydrase IX

CDX2

Caudal type homeobox 2

CEA

Carcinoembryonic antigen

CK7

Cytokeratin 7

CK20

Cytokeratin 20

CPS

Combined Positive Score

DNA

Deoxyribonucleic acid

ER

Estrogen receptor

ECA

Endocervical adenocarcinoma

FFPE

Formalin-fixed paraffin-embedded

GAS

Gastric-type cervical adenocarcinoma

HE

Hematoxylin‒eosin

HNF1β

Hepatocyte nuclear factor 1 beta

HPF

High power field

HPV

Human papillomavirus

HPVA

HPV-associated adenocarcinoma

HPVIA

HPV-independent adenocarcinoma

LNM

Lymph node metastasis

LVSI

Lymphovascular invasion

MMR

Mismatch repair

MSS

Microsatellite stable

MUC6

Mucin 6

NGS

Next generation sequencing

OS

Overall survival

p16

Cyclin-dependent kinase inhibitor 2A

p53

Tumor protein p53

PAX2

Paired box 2

PAX8

Paired box 8

PD-L1

Programmed death-ligand 1

PFS

Progression-free survival

PR

Progesterone receptor

SATB2

Special AT-rich sequence-binding protein 2

TCT

ThinPrep cytology test

TMB

Tumor mutational burden

WHO

World Health Organization

Authors’ contributions

S.G. made substantial contributions to conception and design of study, acquisition of data, drafting of the manuscript and critical revision of the manuscript. Y.S. supervised the study involved in conception and critical revision of the manuscript. All authors read, revise and gave approval of the manuscript.

Funding

There was no funding available for this manuscript.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

The National Cancer Center/Cancer Hospital Ethics Committee, Chinese Academy of Medical Sciences, and Peking Union Medical College approved this study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.