Clinico-pathological features

A total of 74 patients were included in this study, 34 (45.9%) females and 40 (54.1%) males (male to female ratio = 1.2:1), with a median age at cancer diagnosis of 66.5 years (range, 29–89). The majority of cases were of antro-angular localization (58.1%). The most frequent histotype was the poorly cohesive (45.9%) followed by the mixed (37.8%) histotype. Poorly cohesive G/EGJ were classified as PC-NOS (n = 19), PC-SRC (n = 9) and PC-NOS/SRC (n = 6) [18]. Most cases were pT4a (55.7%) for tumor extent and pN3b (29.0%) for lymph node involvement. Fourteen patients (18.9%) received neoadjuvant therapy. The clinico-pathological features of the study cohort are reported in Supplementary Table 1.

Molecular features and expression of predictive biomarkers of primary gastric and esophago-gastric junction carcinomas with peritoneal dissemination

Overall, 50.0% of primary G/EGJ had aberrant p53 expression, showing either a clonal (n = 31) or null (n = 6) phenotype. E-Cadherin was lost in 18.9% of cases, encompassing 13 poorly cohesive carcinomas and the poorly cohesive component of one mixed carcinoma.

As for established predictive biomarkers, only 4.1% of cases were MMRd and 5.4% were HER2-overexpressed (score 3 +). PD-L1 was expressed (CPS ≥ 1) in 79.7% of primary G/EGJ. However, only 14.9% were CPS ≥ 10. As for emerging biomarkers,14.9% of cases were HER2-low, 31.1% were CLDN18 positive and only 1.4% were EBV-associated. Of note, no difference in biomarker staining pattern between the two components was observed in G/EGJ of mixed histotypes.

We used two previously published algorithms [19] for approximating The Cancer Genome Atlas (TCGA) subtypes and Asian Cancer Research Group (ACRG) subtypes using IHC/ISH and morphology. According to TCGA classification, primary tumors were classified as Genomically Stable (GS) in 33 cases (44.6%), Chromosomally Instable (CIN) in 37 cases (50.0%), Microsatellite Instable (MSI) in 3 cases (4.1%) and EBV-associated in one case (1.4%). When using the ACRG classification, primary tumors were classified as Microsatellite Stable/Epithelial Mesenchymal Transition (MSS/EMT) in 33 cases (44.6%), Microsatellite Stable/TP53 active (MSS/TP53 +) in 21 cases (28.4%), Microsatellite Stable/TP53 inactive (MSS/TP53-) in 17 cases (23.0%), and MSI in 3 cases (4.1%).

The molecular features and expression of predictive biomarkers of primary gastric and esophago-gastric junction carcinomas with PM are reported in Table 1 and shown in Fig. 1A.

Table 1 Molecular features and expression of predictive biomarkers in primary gastric and esophago-gastric junction carcinomas (n = 74) with peritoneal disseminationFig. 1

A Histotype and status of predictive biomarkers in gastric and esophago-gastric junction carcinomas with paired peritoneal metastases and B combinations of concordant and discordant biomarkers in each patient (n = 74). Cases with both synchronous and metachronous peritoneal metastases (PM) were considered only once; when the expression differed between synchronous and metachronous (PM), the discordant case was considered. CPS Combined Positive Score, CLDN18 Claudin18, EBER Epstein–Barr virus–encoded small RNA, PD-L1 programmed cell death ligand 1, MMRd mismatch repair deficient, MMRp mismatch repair proficient

Intratumoral heterogeneity of CLDN18 and HER2 in primary G/EGJ

Among the 23 CLDN18-positive primary G/EGJ cases, the majority (16/23; 69.6%) exhibited a homogeneous expression pattern. The remaining 7 cases (30.5%) demonstrated heterogeneous expression. Of these, five cases displayed a random pattern of heterogeneity, one case showed an invasive-front pattern, and in one case of adenosquamous carcinoma, CLDN18 expression was restricted to the adenocarcinoma component.

Among the four HER2-overexpressed primary G/EGJ cases, one exhibited a homogeneous expression pattern, while the remaining three displayed heterogeneous patterns. Of these, two cases had a random heterogeneity pattern, and one case showed superficial positivity.

Comparison of morphology, molecular features and expression of predictive biomarkers between primary gastric and esophago-gastric junction carcinomas and paired PM

Overall, synchronous PM were available for analysis in 60 cases, while metachronous PM were available for analysis in 10 cases. Additionally, in 4 cases, both synchronous and metachronous PM were present.

As for histotype, similar discordance rates were found for synchronous PM (21.9%) and for metachronous PM (28.6%). Discordance was observed only in primary tumors with mixed or mucinous histotype. In 15 out of 28 primary mixed GCs only the PCC (11/15) or the tubular component (4/15) metastasized to the peritoneum. In one patient, the mixed histotype was retained in the synchronous PM, but the tubular component only was found in the metachronous PM. In 2 out of 3 primary tumors of mucinous histotype, only the tubular (1/2) or the PCC (1/2) component metastasized.

An almost perfect concordance between primary tumors and paired PM was observed for E-Cadherin and p53, with an overall discordance rate between primary tumor, synchronous and/or metachronous PM of 5.4% and 1.4% for E-cadherin expression and p53 status, respectively.

A conversion from HER2-overexpressed primary tumor to HER2 not overexpressed metastasis was observed in 1/64 synchronous and 1/14 metachronous PM, with an overall discordance rate of 2.7% (2/74 patients). When scoring HER2 as zero/low/high, the overall discordance rate raised to 27.0% (20/74 patients), with 17/64 discordant synchronous PM and 4/14 discordant metachronous PM. All metachronous PM were HER2-zero. No statistically significant association was found between the homogeneous versus heterogeneous HER2 staining pattern and negative conversion of PM. Among the two primary HER2-overexpressed tumors that underwent negative conversion in PM, one exhibited a homogeneous staining pattern, while the other displayed a heterogeneous staining pattern.

Regarding PD-L1 expression, the prevalence of negative cases (CPS < 1) increased from 20.3% in primary tumors to 29.7% and 28.6% in distant synchronous and metachronous PM, respectively. The overall discordance rate found considering CPS < 1, 1 ≤ CPS < 10, and CPS ≥ 10 as thresholds was 32.4% (24/74 patients), with negative conversion occurring more frequently than positive conversion. The discordance rate for synchronous and metachronous PM was 28.1% and 50.0%, respectively, but no significant difference was found in the number of discordant cases (p-value = 0.13). When considering CPS ≥ vs < 1, the discordance between primary G/EGJ and PM was 23.0%. All PM were concordant with the primary tumor for both MMR and EBER status. Changes in PD-L1 expression between primary tumors and paired PM are shown in Fig. 2A.

Fig. 2

A PD-L1 and B CLDN18 expression in primary tumors (n = 74) and paired peritoneal metastases (PM). Cases with both synchronous and metachronous PM were considered only once; when the expression differed between synchronous and metachronous PM, the discordant case was considered. PD-L1 programmed cell death ligand 1, CPS Combined Positive Score, CLDN18 Claudin18

Comparison of CLDN18 expression between primary gastric and esophago-gastric junction carcinomas and paired PM, and correlation with HER2 and PD‑L1 expression

CLDN18 was positive in 23/74 (31.1%) primary tumors and in 24/78 PM (30.8%; 32.8% in synchronous PM and 21.4% in metachronous PM). No statistically significant difference concerning CLDN18 positivity was found between primary tumors and PM and between synchronous and metachronous PM. In addition, CLDN18 was positive in 31.4% of patients with HER2-not overexpressed primary tumors and in 28.6% of patients with HER2-not overexpressed and PD-L1 CPS < 1 (i.e., double-negative) primary tumors.

The discordance rate for synchronous and metachronous PM was 12.5% and 14.3%, respectively. The overall discordance rate between primary tumor, synchronous and/or metachronous PM was 13.5%, with 5/74 cases each of positive and negative conversion. Changes in CLDN18 expression between primary tumors and paired PM are shown in Fig. 2B. As for HER2, no statistically significant association was found between the homogeneous versus heterogeneous CLDN18 staining pattern and negative conversion of PM. Among the five primary CLDN18-overexpressed tumors that underwent negative conversion in PM, one exhibited a heterogenous staining pattern, while the other four displayed a heterogeneous staining pattern.

Table 2 summarizes the differences in morphology, molecular features and expression of predictive biomarkers between primary tumors and paired synchronous and metachronous PM.

Table 2 Comparison of morphology, molecular features and expression of predictive biomarkers between 74 primary gastric and esophago-gastric junction carcinomas and paired synchronous and metachronous peritoneal metastasesComparison of morphology, molecular features and expression of predictive biomarkers in patients with both synchronous and metachronous PM

Regarding patients with both synchronous and metachronous PM, a complete concordance between primary tumors and both synchronous and metachronous PM was found for MMR, E-cadherin, CLDN18, p53, EBER and HER2 scored as overexpressed/ not overexpressed.

The histotype was concordant between the primary tumor and both PM in 2 cases, while in one case it was mixed in the primary tumor and PC-NOS in both PM and in another one it was discordant only in the metachronous PM (conversion from mixed to tubular histotype).

In two patients HER2 scored as zero/low/high was concordant between the primary tumor and the metachronous PM (HER2-zero) and discordant in the synchronous PM (HER2-low), while in one case the primary tumor was HER2-zero and both metastases were HER2-low.

In one patient the CPS was ≥ 1 and < 10 in the primary tumor and ≥ 10 in both PM.

Implications of neoadjuvant therapy

Of the patients included in the study, 14/74 (18.9%) received neoadjuvant therapy. Among these, one case was pT1 for tumor extent, while all the other cases were pT3 or higher. None of the patients who received neoadjuvant therapy showed discordance for CLDN18 and HER2 expression, while 5/14 were discordant for PD-L1 expression. In 4/14 cases there was a conversion from 1 ≤ CPS < 10 to CPS < 1, and in 1/14 there was a conversion from 1 ≤ CPS < 10 to CPS ≥ 10. However, no significant difference was found in discordance rates between patients who received neoadjuvant therapy and those who did not.

Combinations of concordant and discordant biomarkers

In most patients, (42/74, 56.8%) all predictive biomarkers (i.e., HER2, EBER, PD-L1, CLDN18 and MMR) were concordant between primary tumors and paired PM. In 28/74 patients (37.8%) one predictive biomarker was discordant, while 4/74 patients (5.4%) had two discordant predictive biomarkers. Combinations of concordant and discordant biomarkers found in each patient are shown in Fig. 1B. Examples of discordant biomarkers between primary tumor and peritoneal metastasis are shown in Fig. 3.

Fig. 3

Representative images of discordant biomarkers’ status between primary tumor and matched peritoneal metastasis: A Claudin 18 (CLDN18)-negative primary tumor and B CLDN18-positive peritoneal metastasis, C HER2-overexpressed (3 +) primary tumor and D HER2-not overexpressed (0) peritoneal metastasis, E PD-L1 Combined Positive Score (CPS) 90 in primary tumor and F PD-L1 CPS 0 in peritoneal metastasis