Clinicopathological analysis of claudin 18.2 focusing on intratumoral heterogeneity and survival in patients with metastatic or unresectable gastric cancer Authors: T-Y Kim, Y Kwak, SK Nam, D Han, D-Y Oh, S-A Im, HS Lee Published: 2024 ABSTRACT AbstractBackgroundThis study aimed to investigate the prevalence of claudin 18.2 (CLDN18.2) positivity, with a particular focus on intratumoral heterogeneity, and its association with clinicopathological features in metastatic or unresectable gastric cancer (GC).Patients and methodsWe investigated 400 patients who received systemic chemotherapy for unresectable, metastatic, or recurrent GC. Immunohistochemistry for CLDN18 (43-14A), human epidermal growth factor receptor 2 (HER2), programmed death-ligand 1 (PD-L1), and fibroblast growth factor receptor 2, as well as HER2 silver in situ hybridization (ISH), Epstein–Barr virus (EBV) ISH, and microsatellite instability testing were carried out. CD3+, CD8+, CD4+, and Foxp3-positive immune cell densities were calculated using digital image analysis.ResultsIn GC cases with any CLDN18.2 expression, more than half of the cases (61.3%) showed different expression results between four different tissue microarray (TMA) cores. When comparing CLDN18.2 status between whole tissue sections and the combined results from the four TMA cores, discrepancies were observed in only 2 out of 85 GC cases (2.4%), with 1 false positive and 1 false negative. After considering intratumoral heterogeneity, a CLDN18.2 positivity rate of 31.3% was observed among the 400 GC patients. CLDN18.2 positivity was rare in GCs located in the antrum (or lower third) and in HER2-positive cases but was common in EBV-positive GCs (P < 0.05). No differences in overall survival (OS) were observed according to CLDN18.2 positivity (P = 0.116). Additionally, there was no association between OS and CLDN18.2 positivity in patients treated with fluoropyrimidine plus platinum, chemotherapy plus trastuzumab, paclitaxel with or without ramucirumab, and immuno-oncologic agents. CLDN18.2-positive/PD-L1-high GCs showed statistically significantly longer OS than others (P = 0.025) and higher CD8+ T-cell densities in both the tumor center and periphery (P < 0.001).ConclusionsCharacterizing unresectable, metastatic, or recurrent GC with positive CLDN18.2 expression and evaluating intratumoral heterogeneity and prognostic implications of various therapeutics help advance treatment strategies and develop new therapies for patients with GC.Key words: gastric cancer, claudin 18.2, immunohistochemistry, intratumoral heterogeneity, immune microenvironment Introduction Gastric cancer (GC) remains one of the most lethal malignancies worldwide, with a high incidence and poor prognosis, particularly in advanced stages. Despite advances in diagnostic and therapeutic approaches for various cancers, the first-line treatment targets for advanced GC remain limited to human epidermal growth factor receptor 2 (HER2) and programmed death-ligand 1 (PD-L1). Following the success of the TOGA trial (NCT01041404) in patients with advanced GC in 2010, the HER2 inhibitor trastuzumab was adopted as the first-line treatment.1 For first-line treatment, a combination of immunotherapy, such as nivolumab and pembrolizumab, and chemotherapy has been approved for patients with GC using PD-L1 immunohistochemistry (IHC) as a companion diagnostic in Korea.2, 3, 4 However, the expression of claudin 18.2 (CLDN18.2) in GC has recently garnered significant attention, particularly in light of promising results from phase III clinical trials involving an isoform-specific monoclonal antibody, zolbetuximab, that targets the transmembrane protein CLDN18.2.5, 6, 7 The identification of CLDN18.2 as a potential biomarker has introduced new possibilities for targeted therapy in GC. Ongoing research into the role of CLDN18.2 is essential for refining treatment strategies and improving patient outcomes. The global phase III GLOW and SPOTLIGHT trials demonstrated that zolbetuximab combined with chemotherapy as first-line treatment prolonged progression-free survival in patients with HER2-negative CLDN18.2-positive GC or gastroesophageal junction (GEJ) cancer.5,6 As interest in CLDN18.2 in GC has grown, several single-institution studies have reported the prevalence and clinical relevance of CLDN18.2-positive GC, with variable results due to differences in testing protocols and interpretation of CLDN18.2 IHC.8, 9, 10 Additionally, given the advancement of immunotherapy in GC, the association between CLDN18.2 expression and the immune microenvironment, including PD-L1 expression and tumor-infiltrating T-cell density, needs to be clarified. Comprehensive real-world biomarker data are essential to better understand the role of CLDN18.2 in these contexts and to identify patients who may benefit from targeted therapies, such as zolbetuximab. GC is known to have significant intratumoral heterogeneity in both histological and molecular aspects, which complicates its effective diagnosis and treatment.11 Intratumoral heterogeneity of the histological phenotype is striking in GC, and mixed adenocarcinoma was reported in 6%-22% of cases.12 Our previous study demonstrated a high incidence of intratumoral heterogeneity in HER2 overexpression and gene amplification, which was associated with diffuse or mixed Lauren type and low HER2 expression or amplification.13 Similar to HER2, heterogeneous PD-L1 staining by IHC has been reported in several studies14,15; thus, the intratumoral heterogeneity of CLDN18.2 warrants attention. However, there are currently insufficient data on the variability of CLDN18.2 expression within tumors, particularly in the context of unresectable, metastatic, or recurrent GC. Understanding the heterogeneity of CLDN18.2 expression is crucial, as it may influence therapeutic responses and outcomes. In this study, we aimed to address this gap by focusing on a cohort of patients with unresectable, metastatic, or recurrent GC who were candidates for first-line treatment. Our objectives were to investigate the positivity of CLDN18.2 as defined in clinical trials, to explore its association with crucial GC biomarkers, and to clarify its association with clinicopathological features and the tumor immune microenvironment. In particular, we focused on the intratumoral heterogeneity of CLDN18.2 expression and compared the prognoses of patients with advanced GC receiving different real-world treatments. Patients and methods Patients and clinicopathological information A total of 400 patients treated with first-line chemotherapy for metastatic, recurrent, or unresectable advanced gastric or GEJ adenocarcinoma at Seoul National University Hospital were investigated in this study. Patients who underwent radiotherapy or who did not receive chemotherapy after their diagnosis of advanced disease were excluded. Clinicopathological features, including age, sex, histological type and grade according to the World Health Organization (WHO) classification of tumors, fifth edition, histological classification according to Lauren, gastric tumor location, and staging based on the American Joint Committee on Cancer staging manual eighth edition at initial diagnosis, were documented by reviewing electronic medical records and pathology reports. Details regarding the tissue sample type and anatomical location were extracted from the same sources. The regimens of first-, second-, and third-line chemotherapy and the response to treatment were recorded by reviewing electronic medical charts. Overall survival (OS) was defined as the time from the date of diagnosis of metastatic or unresectable disease to death. The OS of second-line paclitaxel with or without ramucirumab was defined as the time from the start of chemotherapy to the date of death. This study was approved by the Institutional Review Board of Seoul National University Hospital (reference: H-2010-179-1171) and was carried out in accordance with the Declaration of Helsinki for biomedical research involving human subjects. Tissue sample selection and tissue microarray construction Of the 400 advanced GC cases, 286 surgically collected samples were used to construct tissue microarray (TMA) blocks (Superbiochips Laboratories, Seoul, Korea) (Figure 1).16 Four representative cores (three from the tumor center, TMA cores A, B, and C; one from the invasive border, TMA core D) with a diameter of 2 mm were obtained from the original formalin-fixed paraffin-embedded (FFPE) block in each case and rearranged into new TMA blocks as described in our previous publication.16 Specifically, three cores were obtained from the tumor center and one was from the invasive border. The remaining 114 cases involved small biopsy specimens; thus, whole-section slides were used for further analyses. In addition, among 286 surgically collected and TMA-constructed cases, whole sections were additionally prepared in 85 GC cases to compare CLDN18.2 expression status between whole tissue sections and TMA slides. Therefore, whole-section slides were used in 199 GC cases (85 surgically collected and 114 biopsy samples). Schematic of the study design. TMA construction (four cores in each case) was carried out in 286 surgically resected samples, and whole-section slides were used in 114 biopsy samples. Among 286 surgically resected samples, immunohistochemistry on whole-section slides was additionally done in 85 cases to compare between TMA and whole-section slides. CLDN18.2, claudin 18.2; GC, gastric cancer; TMA, tissue microarray. IHC for GC biomarkers IHC was conducted using the VENTANA (43-14A) Class 1 IVD Assay (43-14A, ready to use; Ventana Medical Systems, Tucson, AZ) using a Ventana BenchMark ULTRA autostainer (Ventana Medical Systems). IHC for CLDN18 was carried out using normal gastric mucosa as internal positive control and stromal cells, endothelial cells, fibroblasts, and immune cells as internal negative controls to ensure the quality of CLDN18 IHC staining. The staining intensity was scored as negative (0), weak (1), moderate (2), or strong (3), and the positive area was recorded as a percentage. Moderate-to-strong membranous staining in ≥75% of tumor cells was defined as CLDN18.2 positivity.17 IHC for other GC biomarkers was carried out using HER2 (4B5, prediluted, Ventana Medical Systems) and fibroblast growth factor receptor 2 (FGFR2) (SP273, 1 : 100, Abcam, Cambridge, UK) using a Ventana BenchMark ULTRA autostainer (Ventana Medical Systems). HER2 and FGFR2 positivity was assessed by applying an area cut-off of at least 10% stained tumor cells for resection specimens, and for biopsy specimens, positivity was defined as the presence of a small single cluster of cells (or at least five cells) showing membranous staining. Membranous staining was graded as 1+ for faint staining, 2+ for weak-to-moderate staining, and 3+ for strong staining.18 FGFR2 positivity was defined as 2+ or 3+, as previously described.19,20 PD-L1 (22C3 pharmDx assay, Dako, Carpinteria, CA) IHC was carried out using a Dako autostainer Link 48 (Dako). The combined positive score (CPS) of PD-L1 staining was recorded according to the 22C3 pharmDx interpretation guidelines.21 For survival analysis, PD-L1 staining status was divided into high (CPS ≥ 5) and low (CPS < 5) levels. IHC for immune cell densities To obtain tumor-infiltrating immune cell densities, we carried out IHC for CD3 (UCHT1, 1 : 300, Dako), CD8 (C8/144B, prediluted, Dako), CD4 (SP35, prediluted, Ventana Medical Systems), and Foxp3 (236A/E7, 1 : 50, Abcam) on two TMA cores (one from the tumor center and one from the tumor periphery) in 201 GC cases.22 IHC slides were scanned using Aperio AT2 (Leica Biosystems, Wetzlar, Germany) at ×40 magnification, and an experienced pathologist (HSL) annotated the tumor area in each core. CD3+, CD8+, CD4+, and Foxp3-positive cell densities per mm2 were calculated for each annotated area using QuPath v.0.3.1. Microsatellite instability testing The tumor and paired non-neoplastic gastric mucosa were macrodissected from the FFPE blocks, and genomic DNA was extracted. To assess microsatellite instability (MSI) status, two mononucleotide markers (BAT25 and BAT26) and three dinucleotide markers (D2S123, D5S346, and D17S250) were used according to the National Institutes of Health guidelines. After PCR amplification, fragmentation analysis was carried out using an ABI 3731 automatic DNA sequencer (Applied Biosystems, Foster City, CA). MSI-high (MSI-H) was defined as a tumor with two or more unstable markers, MSI-low was defined as a tumor with one unstable marker, and microsatellite stability was defined as the absence of an unstable marker.23 Epstein–Barr virus in situ hybridization Epstein–Barr virus (EBV) in situ hybridization (ISH) was carried out using an automated EBV-encoded RNA ISH. Tissue sections were digested and hybridized using a BOND EBER probe (Leica Microsystems, Wetzlar, Germany). The positivity of the tumor cell nuclei was assessed. HER2 silver ISH Dual-color HER2 silver ISH (SISH) was carried out using an automatic SISH staining device (BenchMark ULTRA) according to the manufacturer’s instructions for the INFORM HER2 DNA and INFORM Chromosome 17 (CEP17) probes (Ventana Medical Systems). In each case, a total of 20 tumor cells were counted, and a HER2/CEP17 ratio ≥2.0 was defined as HER2 amplification. HER2 positivity was defined as HER2 IHC 3+ or IHC 2+ with HER2 amplification.21 Statistical analysis The association between CLDN18.2, clinicopathological features, and GC biomarkers was analyzed using the chi-square or Fisher’s exact test for categorical variables, Student’s t-test or analysis of variance for parametric tests of continuous variables, and the Mann–Whitney test or Kruskal–Wallis test for non-parametric tests of continuous variables. Pairwise comparison of CLDN18.2 status between whole-section slides and TMA cores was carried out by calculating the overall percent agreement (OPA), positive percent agreement (PPA), and negative percent agreement (NPA). The Fleiss κ value for agreement between the TMA cores was evaluated. Kaplan–Meier survival curves were plotted for OS analysis, and P values were calculated using the log-rank test. Estimates of medians and 95% confidence intervals were derived for each sample group for the exploratory analysis. Statistical analyses and visualizations were carried out using R version 4.4.1. Statistical significance was set at P < 0.05. Results Patient demographics A total of 400 patients with metastatic or unresectable GC who received palliative chemotherapy were included in this study. Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2024.104000, shows the clinicopathological features of the study population. Among the 400 patients, 180 (45.0%) were initially diagnosed with stage IV GC with distant metastases, 23 (5.8%) were initially unresectable, 15 (3.8%) underwent R1 or R2 resection, and 182 (45.5%) were diagnosed with metastatic or recurrent disease after radical gastrectomy. We collected 286 surgically resected samples (gastrectomy before chemotherapy, n = 252; gastrectomy after chemotherapy, n = 30; and metastasectomy before chemotherapy, n = 4) and 114 biopsy samples (endoscopic/needle/punch biopsy before chemotherapy, n = 98; endoscopic/needle/punch biopsy after chemotherapy, n = 2; small excisional biopsy before chemotherapy, n = 2; and small excisional biopsy after chemotherapy, n = 2). Fluoropyrimidine plus platinum (FP) combination therapy (312, 78.0%) was the most commonly used first-line treatment, followed by chemotherapy plus trastuzumab (35, 8.8%) and FP plus immuno-oncologic agents (IO) (22, 5.5%) (Supplementary Figure S1, available at https://doi.org/10.1016/j.esmoop.2024.104000). Paclitaxel with or without ramucirumab was the most frequently used second-line treatment, whereas irinotecan-based chemotherapy was commonly administered as a third- or later-line treatment. Fifty-four (13.5%) patients were exposed to IO regardless of the line of treatment. All 400 patients received first-line treatment, with 212 (53.0%) receiving second-line, 92 (23.0%) receiving third-line, and 47 (11.8%) receiving fourth-line or later treatment. Intratumoral heterogeneity of CLDN18.2 expression in advanced GC To evaluate the intratumoral heterogeneity of GC, four different areas of each primary tumor were examined by preparing TMA blocks with four representative cores (A, B, C, and D in each case) from 286 surgically resected cases (Figure 1). Analysis revealed that CLDN18.2 positivity varied across the cores: A, 28.3%; B, 28.7%; C, 26.6%; and D, 27.3% (Figure 2A). Specifically, 175 patients (61.2%) showed no CLDN18.2 positivity in any core, whereas 111 patients (38.8%) exhibited CLDN18.2 positivity in one or more cores. Among 111 cases with CLDN18.2 positivity in one or more cores, homogeneous expression of CLDN18.2 (positive in all four cores) was observed in 43 cases (38.7%), and the remaining 68 cases (61.3%) had heterogeneous expression of CLDN18.2 across TMA cores from the same tumor (positive in one, two, or three cores; 30, 17, and 21 cases, respectively) (Figure 2B). CLDN18.2 positivity in individual TMA cores and intratumoral heterogeneity. (A) CLDN18.2 positivity rates in each single TMA core (cores A, B, C, and D), as well as the positivity rate in at least one of these cores (positive in ≥1 core). (B) The results of CLDN18.2 immunohistochemistry in 111 cases where CLDN18.2 was positive in at least one core. Homogeneous positivity in all four TMA cores was observed in 43 cases (38.7%) and heterogeneous positivity (different results between TMA cores) occurred in the remaining 68 cases (61.3%). CLDN18.2, claudin 18.2; TMA, tissue microarray. Heterogeneous expression of CLDN18.2 was observed in multiple TMA cores. Therefore, to reveal the accuracy of TMA cores, we compared CLDN18.2 positivity between TMA and whole tissue sections in 85 gastrectomy samples. The analysis revealed discordance between CLDN18.2 status in a single TMA core and whole sections in 7 of 85 cases (8.2%) in core A, 6 (7.0%) in core B, 4 (4.8%) in core C, and 8 (9.4%) in core D (Supplementary Table S2, available at https://doi.org/10.1016/j.esmoop.2024.104000). There were no significant differences between the false-positive (positive in single TMA core, but negative in whole section) and false-negative (negative in single TMA core, but positive in whole section) rates of results from individual TMA cores when whole-section results were considered the gold standard. OPA, PPA, and NPA ranged from 90.6% to 95.3%, 85.7% to 92.9%, and 91.2% to 96.5%, respectively. In contrast to the results from a single TMA core, only 2 out of 85 GCs (2.4%) showed discordant results between whole tissue sections and the combined analysis of all four TMA cores, i.e. area of moderate-to-strong membranous expression of CLDN18.2 was calculated across all four TMA cores (Supplementary Table S2, available at https://doi.org/10.1016/j.esmoop.2024.104000). In one case, CLDN18.2 was positive in the combined analysis of all four TMA cores (false positive), but negative in the whole section (Supplementary Figure S2A, available at https://doi.org/10.1016/j.esmoop.2024.104000). In contrast, the opposite was found in another case (false negative) (Supplementary Figure S2B, available at https://doi.org/10.1016/j.esmoop.2024.104000). OPA, PPA, and NPA of CLDN18.2 status from the combined analysis of four TMA cores compared with whole sections were 97.7%, 96.4%, and 98.3%, respectively. Based on these findings, CLDN18.2 status determined by the combined analysis of the four TMA cores was used in 286 surgically resected samples with TMA construction for further analysis. CLDN18.2 expression in GC and association with clinicopathological features Of 400 patients with metastatic, recurrent, or unresectable GC receiving systemic chemotherapy, 125 cases (31.3%) were positive for CLDN18.2 IHC (Supplementary Table S3, available at https://doi.org/10.1016/j.esmoop.2024.104000). When comparing the sample location or type used for CLDN18.2 IHC, the positivity rate was 30.5% (105/344) for primary tumor samples (gastrectomy or endoscopic biopsy), 33.3% (10/30) for primary tumor samples after chemotherapy (all gastrectomy samples), and 31.3% (5/16) for seeding tissue samples. The positivity rates were similar among the various sample types, even in small endoscopic/needle/punch biopsy samples (33.3%, 33/100). CLDN18.2 positive expression was less frequently observed in GCs located in the antrum or lower third (Supplementary Table S4, available at https://doi.org/10.1016/j.esmoop.2024.104000). When compared with histological type based on the WHO classification, CLDN18.2 positive expression was rare in mucinous adenocarcinoma (1/9, 11.1%). Although there was only one case of gastric carcinoma with lymphoid stroma in our cohort, this case showed positive CLDN18.2 expression. The CLDN18.2 positivity rate of intestinal type by Lauren (53/150, 35.3%) was not significantly higher than that of diffuse- (65/225, 28.9%) or mixed-type GCs (7/25, 28.0%) (P = 0.392). CLDN18.2-positive expression was not associated with other clinicopathological features, including age, sex, and TNM (tumor–node–metastasis) stage at initial diagnosis. Survival analysis according to CLDN18.2 expression status CLND18.2 positivity was not significantly associated with OS in the overall study population (17.9 months in CLDN18.2-negative GC versus 19.2 months in CLDN18.2-positive GC, P = 0.116) (Figure 3A, Supplementary Table S5, available at https://doi.org/10.1016/j.esmoop.2024.104000). In patients treated with FP as the first-line regimen, there were no significant differences in OS based on CLDN18.2 status (17.9 versus 19.2 months, P = 0.118) (Figure 3B). Similarly, OS did not differ significantly between patients with CLDN18.2-positive and -negative GC who received chemotherapy plus trastuzumab as first-line treatment (Figure 3C), paclitaxel with or without ramucirumab as second-line therapy (Figure 3D), or those who were exposed to IO in any line (Figure 3E). Additionally, there was no significant difference in OS between patients who received irinotecan-based chemotherapy in the third or later lines (Figure 3F). OS analysis based on treatment regimen. (A) OS between CLDN18.2-negative and CLDN18.2-positive patients across the entire study population, regardless of treatment. (B) OS analysis of patients receiving first-line fluoropyrimidine plus platinum chemotherapy. (C) Comparison of OS in patients treated with trastuzumab plus chemotherapy. (D) OS in patients treated with second-line paclitaxel with or without ramucirumab. (E) OS in patients who were exposed to IO. (F) OS in patients treated with irinotecan-based third- or later-line treatment. (G-H) Analysis of OS in patients not exposed to IO. (G) Stratified by CLDN18.2 and PD-L1 status. (H) Survival difference between CLDN18.2-positive/PD-L1-high GC and all other groups (PD-L1 high defined as PD-L1 IHC CPS ≥5). CI, confidence interval; CLDN18.2, claudin 18.2; CPS, combined positive score; CTx, chemotherapy; IHC, immunohistochemistry; IO, immuno-oncologic agents; NA, not available; OS, overall survival; PD-L1, programmed death-ligand 1. Notably, in patients not exposed to IO (n = 346), the group with both positive CLDN18.2 expression and high PD-L1 levels (PD-L1 CPS ≥ 5) exhibited a trend toward longer OS (Figure 3G, Supplementary Table S5, available at https://doi.org/10.1016/j.esmoop.2024.104000). When comparing the OS between this group (positive CLDN18.2 and high PD-L1) and all other groups, the former demonstrated a more prolonged OS (20.1 versus 16.5 months, P = 0.025) (Figure 3H). However, after further excluding patients with positive EBV and MSI-H (n = 31), while the difference in OS between the two groups was maintained, it did not reach statistical significance (24.6 versus 16.5 months, P = 0.166) (Supplementary Table S5, available at https://doi.org/10.1016/j.esmoop.2024.104000). Relationship of CLDN18.2 positivity with other GC biomarkers and immune microenvironment We compared CLDN18.2 expression status with that of other GC biomarkers, including FGFR2, HER2, MSI, and EBV (Table 1). CLDN18.2-positive GC tended to be HER2 negative (P = 0.048) and EBV positive (P = 0.001). There was no statistically significant association between CLDN18.2 and FGFR2 or MSI status (P > 0.05). Additionally, we compared CLDN18.2 expression patterns (negative, heterogeneous, and homogeneous) and observed that EBV and PD-L1 positivity rates were significantly higher in cases with homogeneous CLDN18.2 expression than in those with heterogeneous expression (P < 0.05) (Supplementary Table S6, available at https://doi.org/10.1016/j.esmoop.2024.104000). Table 1Correlation between CLDN18.2 expression status and other GC biomarkersCLDN18TotalP valueNegativePositiveFGFR2Negative237993360.07770.5%29.5%84.0%Positive38266459.4%40.6%16.0%HER2Negative2451193640.04867.3%32.7%91.0%Positive3063683.3%16.7%9.0%MSIMSS/MSI-L2591133720.16969.6%30.4%93.0%MSI-H16122857.1%42.9%7.0%EBVNegative2701143840.00170.3%29.7%96.0%Positive5111631.3%68.8%4.0%PD-L1 (cut-off: CPS 1)Negative99431420.75769.7%30.3%35.5%Positive1768225868.2%31.8%64.5%PD-L1 (cut-off: CPS 5)Negative193832760.44869.9%30.1%69.0%Positive824212466.1%33.9%31.0%PD-L1 (cut-off: CPS 10)Negative224993230.59669.3%30.7%80.8%Positive51267766.2%33.8%19.3%Total27512540068.8%31.3%100.0%CLDN18.2, claudin 18.2; CPS, combined positive score; EBV, Epstein–Barr virus; FGFR2, fibroblast growth factor receptor 2; GC, gastric cancer; HER2, human epidermal growth factor receptor 2; MSI, microsatellite instability; MSI-H, MSI-high; MSI-L, MSI-low; MSS, microsatellite stable; PD-L1, programmed death-ligand 1. Correlation between CLDN18.2 expression status and other GC biomarkers CLDN18.2, claudin 18.2; CPS, combined positive score; EBV, Epstein–Barr virus; FGFR2, fibroblast growth factor receptor 2; GC, gastric cancer; HER2, human epidermal growth factor receptor 2; MSI, microsatellite instability; MSI-H, MSI-high; MSI-L, MSI-low; MSS, microsatellite stable; PD-L1, programmed death-ligand 1. To investigate the immune microenvironment of CLDN18.2-positive GCs, PD-L1 IHC using the 22C3 pharmDx assay and IHC for CD3, CD8, CD4, and Foxp3 were carried out. PD-L1 positivity at cut-offs of CPS1, CPS5, or CPS10 was slightly higher in CLDN18.2-positive GC than in CLDN18.2-negative GC; however, the difference was not statistically significant (P > 0.05) (Table 1). Tumor-infiltrating T-cell densities were calculated at the tumor center and periphery in each case using digital image analysis. CD3+ T-cell density in the tumor periphery (P = 0.009), CD8+ T-cell density in the tumor center (P = 0.041), and CD4+ T-cell density in the tumor periphery (P = 0.040) were higher in CLDN18.2-positive GCs than in CLDN18.2-negative GCs (Supplementary Figure S3, available at https://doi.org/10.1016/j.esmoop.2024.104000). The immune cell densities were compared according to PD-L1 and CLDN18.2 status (Figure 4). The group with both positive CLDN18.2 expression and high PD-L1 levels (PD-L1 CPS ≥ 5) tended to show higher immune cell densities than other groups, especially CD8+ T cells at the tumor center (P < 0.001) and CD3+, CD4+, and CD8+ T cells at the tumor periphery (P < 0.001). After excluding MSI-H and EBV-positive cases, the group with both positive CLDN18.2 expression and high PD-L1 levels still tended to show higher immune cell densities than other groups, especially CD8+ T cells (Supplementary Figure S4, available at https://doi.org/10.1016/j.esmoop.2024.104000). Additionally, we carried out further analysis using a higher PD-L1 cut-off of CPS ≥10. While the difference in immune cell densities between the PD-L1-high and -low groups became slightly more pronounced with this higher threshold, the overall statistical significance of these differences did not change substantially (Supplementary Figure S5, available at https://doi.org/10.1016/j.esmoop.2024.104000). Immune cell densities according to CLDN18.2 and PD-L1 status (negative + low, CLDN18.2 negative and PD-L1 CPS <5; positive + low, CLDN18.2 positive and PD-L1 CPS <5; negative + high, CLDN18.2 negative and PD-L1 CPS ≥5; positive + high, CLDN18.2 positive and PD-L1 CPS ≥5). CLDN18.2, claudin 18.2; CPS, combined positive score; PD-L1, programmed death-ligand 1. Discussion Following the success of the SPOTLIGHT and GLOW trials, CLDN18.2 has emerged as an important biomarker in patients with advanced GC.5,6 Wide variation in CLDN18.2 positivity has been reported in several previous studies using various detection antibodies and positivity criteria, and furthermore, IHC for CLDN18.2 was carried out on surgically resected samples at various stages in most previous studies.10,24,25 In this study, we investigated patients with metastatic, recurrent, or unresectable GC who received systemic chemotherapy and carried out CLDN18.2 IHC with the antibody (VENTANA 43-14A clone) used in the clinical trials. The definition of CLDN18.2 positivity was the same as that used in pivotal clinical trials. CLDN18.2 was positive in 125 of 400 (31.3%) patients with advanced GC, which was lower than the CLDN18.2 positivity in the SPOTLIGHT and GLOW trials (38.4%). However, the prevalence data from previous studies showed a lower positivity rate of CLDN18.2 in Asian patients than in European or North American patients,26 and a previous study reported a CLDN18.2 positivity rate of 24.0% in Japanese advanced GC.27 Our study provides essential baseline results that may significantly influence future treatment strategies targeting CLDN18.2 in patients with GC. In our study, no specific association between CLDN18.2 expression and OS was observed in all study populations or across various treatment groups, including FP, chemotherapy plus trastuzumab, paclitaxel with or without ramucirumab, IO, and irinotecan-based chemotherapy. To our knowledge, our study is one of the largest studies to evaluate the relationship between CLDN18.2 positivity and OS, especially in stage IV GC, across various treatments. Furthermore, despite the lack of differences in OS according to CLDN18.2 expression, we identified an interesting finding within a specific subgroup of patients with positive CLDN18.2 expression and high PD-L1 levels. These patients showed significantly longer OS than others, suggesting that the combination of CLDN18.2 and PD-L1 could be valuable in refining new treatment strategies. Marked intratumoral and intertumoral heterogeneity is a distinguishing feature of GC.12 In addition to histological heterogeneity, previous studies have reported heterogeneity in HER2 expression and gene amplification,13,28 and PD-L1 expression in GC.14,15 Because of intratumoral heterogeneity, when using endoscopic biopsy samples, at least four biopsies should be collected and analyzed for HER2 and PD-L1 expression.15,28 Similar to HER2 and PD-L1, heterogeneous expression between cores from the same tumor was observed in 68 out of 111 resected GC cases (61.3%) when we compared CLDN18.2 expression across four TMA cores in this study. However, when four TMA cores were analyzed combined, OPA was 97.7%, PPA was 96.4%, and NPA was 98.3% compared with the whole-slide sections. Therefore, a single tissue core is insufficient for CLDN18.2 diagnosis, and multiple cores are required to make an accurate diagnosis for each patient. Additionally, we estimated the false-negative and false-positive rates of TMA cores using whole-slide sections as the gold standard. In a study of PD-L1 IHC, false negatives were more frequent than false positives in biopsy samples.29 Nevertheless, the false-negative and false-positive rates for CLDN18.2 assessment in this study were similar. Our study provides crucial insights into the intratumoral heterogeneity of CLDN18.2 expression in GC, a factor that has not been fully explored in the literature. GC is one of the main targets of immunotherapy using programmed cell death protein 1/PD-L1 inhibitors. Therefore, in addition to assessing CLDN18.2 expression, our study explored its association with the immune response, particularly PD-L1 expression and tumor-infiltrating T-cell densities. In the SPOTLIGHT and GLOW trials, PD-L1 CPS ≥5 in CLDN18.2-positive GC was observed in 17.2% of cases, but the rate of PD-L1 positivity in CLDN18.2-positive GC varies across studies depending on the case number, antibody used, or positivity definition.8,30,31 Our results showed that PD-L1 positivity, defined as CPS ≥1, 5, or 10, did not significantly differ between CLDN18.2-positive and -negative GC, which is consistent with the results of Kubota et al.27 When CD3+, CD4+, CD8+, and Foxp3-positive cell densities were calculated in the tumor center and periphery, CD3+ and CD4+ cells in the tumor periphery and CD8+ cells in the tumor center were more abundant in CLDN18.2-positive GC than in CLDN18.2-negative GC. Furthermore, the combined analysis of CLDN18.2 with PD-L1 showed that CLDN18.2-positive/PD-L1-high GC had significantly higher densities of CD8+ T cells in the tumor center and CD3+, CD4+, and CD8+ T cells in the tumor periphery than the other groups, suggesting the influence of CLDN18.2 on the tumor immune microenvironment. Previous research, such as the study by Jia et al.,31 has demonstrated similar trends in immune infiltration in CLDN18.2-positive tumors, but the number of enrolled participants was small, and the definition of CLDN18.2 positivity in that study differed from ours. Further comprehensive and large-scale studies on the impact of CLDN18.2 on the tumor immune microenvironment are needed. Our data revealed the clinicopathological features of CLDN18.2-positive GC. CLDN18.2 IHC was rarely positive in GC located in the antrum or lower third (P < 0.05). CLDN18.2 is specifically expressed in normal epithelial cells of the gastric mucosa, and its lower expression has been reported to occur during intestinal metaplasia and gastric carcinogenesis.32,33 Lower expression of CLDN18.2 was more frequently observed in GC with an intestinal phenotype,32 suggesting that CLDN18.2 positivity may be less common in lower-third GC. When comparing CLDN18.2 status with molecular subtypes and several crucial biomarkers in our data, CLDN18.2 positivity was more common in EBV-positive GC (P = 0.001). CLDN18.2 was positive in 68.8% of EBV-positive GC cases. This prevalence is consistent with previous translational studies,31,34 although another study reported 26.7% of CLDN18.2 positivity in EBV-positive GC.27 In contrast to EBV-positive GC, CLDN18.2 positivity was lower in HER2-positive GC than in HER2-negative GC (16.7% versus 32.7%, P = 0.048). Therefore, CLDN18.2 positivity increased in patients with HER2-negative GC, who are actual candidates for zolbetuximab, compared with the overall patients with GC, but the difference was minimal because the HER2 positivity rate in advanced GC is low (9.0%). In addition, following a successful phase II study of an anti-FGFR2b monoclonal antibody in GC, FGFR2b has emerged as a potential biomarker for the therapeutic targeting of GC.35 Therefore, we compared FGFR2 with CLDN18.2 status in advanced GC, but there was no significant association. A limitation of our study was the lack of a comparative analysis between primary and metastatic lesions in the same patients. Such a comparison could provide additional insights into the potential variability in CLDN18.2 expression across different tumor sites, which may affect treatment response and prognosis. Our study also included a relatively small number of post-chemotherapy tumor samples, and patients received various chemotherapeutic regimens. To gain a comprehensive understanding of CLDN18.2 expression in potential candidates, further studies with larger post-treatment cohorts will be needed. Furthermore, the lack of data on zolbetuximab treatment outcomes constrains our ability to fully assess the clinical relevance of CLDN18.2 expression in this context. As more data on this targeted therapy become available, future research will be better positioned to explore these associations. In conclusion, the CLDN18.2 positivity rate was 31.3% in patients with advanced GC eligible for first-line treatment. We emphasize the importance of accounting for intratumoral heterogeneity when assessing CLDN18.2 and that the combination of CLDN18.2 and PD-L1 expression has prognostic significance. In addition, CLDN18.2-positive and PD-L1-high GC showed higher densities of various T cells than the other groups. These findings will help apply CLDN18.2 as a target for GC treatment. Further research is needed to validate these associations and explore their implications in the context of emerging therapies, such as zolbetuximab. Acknowledgments Funding This work was supported by Astellas Pharma Inc., Seoul, Republic of Korea (no grant number). Disclosure The corresponding author HSL received research funding from Astellas Pharma Inc., Seoul, Republic of Korea for this study. The co-author DH is an employee of Astellas Pharma Inc., Seoul, Republic of Korea. The research funder did not play any role in the research design, execution, analysis, interpretation, and reporting. All other authors have declared no conflicts of interest. Data sharing The data used and/or analyzed during the current study are available from the corresponding author on reasonable request. Supplementary data Supplementary Tables Supplementary Figures References Associated Data Supplementary Materials Supplementary Tables Supplementary Figures