Claudin 18.2 Gastric Cancer: Mechanisms and Prognostic Impact

Claudin 18.2 (CLDN18.2) is a protein normally buried within the tight junctions of healthy stomach lining cells, invisible to the immune system and to drugs circulating in the bloodstream. When gastric cells become cancerous, the orderly cell architecture breaks down and CLDN18.2 becomes exposed on the tumor surface, turning it into a target that therapies can reach.¹ Roughly four in ten patients with advanced gastric or gastroesophageal junction cancer carry tumors that express enough CLDN18.2 to qualify for targeted treatment, and that distinction now shapes first-line therapy decisions in ways that were impossible just a few years ago.

How CLDN18.2 Becomes a Target During Malignant Transformation

In a healthy stomach, claudin 18.2 sits sandwiched between adjacent epithelial cells, forming part of the seal that keeps gastric acid on one side and the body’s internal environment on the other. The protein’s outer loops face inward, locked within those junctions, so antibodies and immune cells in the blood have no way to reach them. When a gastric tumor forms, cells lose their normal polarity and tight-junction organization. CLDN18.2 migrates to the exposed outer surface of the cancer cell membrane, where its extracellular domain is now accessible.¹

This exposure is not just a passive side effect of disordered cell growth. Laboratory work has shown that CLDN18.2 on the tumor surface can actively participate in cancer progression. In co-culture experiments, cancer-associated fibroblasts (a type of support cell that tumors recruit) promoted the migration and invasion of gastric cancer cells that expressed CLDN18.2, but not those without it. When researchers knocked down CLDN18.2 expression using targeted gene silencing, the migration advantage disappeared.² The implication is that CLDN18.2 is not simply a bystander marker sitting on the tumor surface. It mediates a real interaction between the tumor and its surrounding stroma that helps the cancer spread.

How Common Is CLDN18.2 Positivity

The answer depends on where the cutoff is set and which population is tested, but the ballpark figure across large screening datasets is that roughly 38 to 47 percent of advanced gastric cancers are CLDN18.2-positive using the clinically relevant threshold. Data pooled from the SPOTLIGHT and GLOW trials, which screened over 4,500 patients using the companion diagnostic assay, found a combined CLDN18.2 positivity rate of 38.4% when requiring moderate-to-strong membrane staining in at least 75% of tumor cells.³ A separate retrospective analysis of 304 tumors using the same cutoff found 44.4% positivity overall, with pure gastric tumors having a higher rate (about 51%) than gastroesophageal junction tumors (roughly 35%).⁴ A study focused on metastatic gastric cancer patients found 47% positivity.⁵

If you relax the threshold to count tumors with at least moderate staining in 50% or more of cells, the numbers climb further: around 64% for gastric adenocarcinoma and 45% for gastroesophageal junction tumors in one cohort.⁴ This gap between cutoffs matters because newer drug types, particularly antibody-drug conjugates, may work at lower expression levels than the threshold approved for zolbetuximab, potentially expanding the eligible patient population.

Which Tumors Are More Likely to Express CLDN18.2

CLDN18.2 positivity is not randomly distributed across all gastric cancers. It clusters with certain clinical and molecular features. Tumors located in the body or fundus of the stomach (non-antral, non-junction sites) tend to express CLDN18.2 more often than those at the gastroesophageal junction.⁶ Diffuse-type gastric cancer, the subtype with signet-ring cells and a sheet-like growth pattern rather than gland formation, shows a stronger association with CLDN18.2 expression than the intestinal type.⁷ Epstein-Barr virus (EBV)-associated gastric cancers stand out particularly, with one large study reporting CLDN18.2 positivity in about 72% of EBV-positive tumors.⁸ A Chinese cohort also found associations with younger age and female sex.⁶

The Relationship With HER2, PD-L1, and Microsatellite Instability

One of the most clinically relevant features of CLDN18.2 is that it tends to appear in tumors that are HER2-negative, precisely the population that has historically had the fewest targeted treatment options. In one study of over 1,000 gastric cancers, only about 18% of HER2-positive tumors co-expressed CLDN18.2, compared with much higher rates in HER2-negative tumors.⁸ Among HER2-negative tumors with peritoneal spread, roughly 29% were positive at the 75% cutoff.⁹ This inverse relationship with HER2 means that CLDN18.2 fills a genuine treatment gap rather than overlapping with existing targeted options.

The picture with PD-L1 is less clean-cut. Some studies report a modest positive correlation. CLDN18.2-positive tumors showed slightly higher PD-L1 positivity across multiple PD-L1 scoring thresholds in one large analysis.⁸ Other studies found no significant correlation at all.⁹ As for microsatellite instability, CLDN18.2 expression appears to be similar in microsatellite-stable and microsatellite-instability-high tumors, with no meaningful difference between the groups.⁸ Practically, this means CLDN18.2 testing yields actionable results across the molecular landscape of gastric cancer rather than being confined to a single molecular subgroup.

What CLDN18.2 Expression Means for Prognosis

Whether CLDN18.2 expression by itself tells you something about how a patient will do, independent of any targeted treatment, is genuinely unsettled. Some data suggest it carries a negative prognostic signal. One study identified CLDN18.2 expression as an independent risk factor for poorer overall survival, with a hazard ratio of about 1.57 in a multivariate analysis adjusted for other clinical variables.¹⁰ But a separate study of 65 advanced patients found no statistically significant link between CLDN18.2 expression and either progression-free or overall survival, with older age being the only independent predictor of poor outcomes in that cohort.¹¹

This disagreement likely reflects the complexity of the question. CLDN18.2-positive tumors tend to cluster with diffuse histology, which itself carries a worse prognosis, so separating the independent contribution of CLDN18.2 from the histologic subtype it associates with is statistically difficult in smaller cohorts. The story also changes depending on the treatment the patient receives. When CLDN18.2-positive tumors are treated with a drug that targets that protein, the biomarker obviously becomes predictive of benefit rather than prognostic in the traditional sense. For clinicians, the more actionable question has shifted from “does CLDN18.2 expression predict survival on its own” to “does it predict who benefits from CLDN18.2-targeted therapy,” and on that second question the evidence is much clearer.

How CLDN18.2 Is Tested and the Problem of Heterogeneity

CLDN18.2 is measured by immunohistochemistry (IHC) on tumor tissue, typically a biopsy or surgical specimen. Pathologists score the staining by intensity (from 0 for no staining up to 3+ for strong staining) and by the percentage of tumor cells that show membrane staining, estimated in 10% increments.¹² The FDA-approved threshold for zolbetuximab eligibility is moderate-to-strong (2+ or 3+) staining in at least 75% of tumor cells.¹³ Lower thresholds are being explored for other drugs, with some investigators evaluating cutoffs at an H-score of 25 or even as low as 10.¹⁴

A major practical challenge is that CLDN18.2 expression is not uniform throughout a tumor. About 39% of primary tumors in one tissue-microarray study showed significant staining variability from one area of the tumor to another.⁷ Intratumoral heterogeneity means a single small biopsy can miss patches of high expression or, conversely, sample an unusually bright spot that does not represent the whole tumor. And the problem extends beyond the primary tumor: when researchers compared primary and metastatic sites in the same patients, roughly 20 to 25% showed discordant results.¹⁵ ¹⁶ The pattern varies by metastatic site: peritoneal metastases tend to retain CLDN18.2 expression (positivity around 44%), while liver metastases are much less likely to be positive (around 18%).¹⁶

Antibody clone selection for the IHC assay also matters. The 43-14A clone, used in the companion diagnostic approved alongside zolbetuximab, has been the most widely validated and consistently produces optimal staining. Proficiency testing across laboratories found that some alternative clones, including ZR451 and EPR19202, were associated with false-negative results in certain settings.¹⁷ That said, head-to-head comparisons of well-validated clones have shown agreement rates above 0.8 by kappa and highly consistent treatment-response predictions, so the concern is mainly about laboratories using less-established reagents.¹⁸ ¹⁹

The Tumor Microenvironment Around CLDN18.2-Positive Tumors

Tumors do not exist in isolation; they are embedded in a complex neighborhood of immune cells, fibroblasts, blood vessels, and signaling molecules. CLDN18.2-positive gastric cancers appear to cultivate a distinct version of this neighborhood. A systematic review found a consistent positive correlation between CLDN18.2 expression and the infiltration of CD8+ T cells (the immune cells most directly responsible for killing cancer cells), neutrophils, and cancer-associated fibroblasts.²⁰ CD4+ T cells and macrophages have also been reported at higher levels in some cohorts.²¹ Individual studies have confirmed that CD4+ and CD8+ T-cell fractions are significantly higher in CLDN18.2-positive tumors than in negative ones.¹⁰

This “hotter” immune microenvironment is a double-edged observation. On one hand, more immune infiltration might make these tumors better candidates for immunotherapy combinations. On the other, the enrichment of cancer-associated fibroblasts, as shown in the co-culture experiments described earlier, suggests these same tumors may have a stromal support network that actively promotes invasion.² Whether the net effect of this microenvironment tilts toward immune control or tumor promotion likely depends on the specific balance of cell types in each patient’s tumor. For the broader field, the observation has motivated trials combining CLDN18.2-targeting agents with checkpoint inhibitors, the logic being that a tumor already infiltrated by T cells might respond more robustly if you also remove the brakes on those T cells.

Zolbetuximab and the First Approved CLDN18.2-Targeted Therapy

Zolbetuximab is a monoclonal antibody that binds to the exposed extracellular domain of CLDN18.2 on tumor cells. It kills those cells through two main immune-mediated mechanisms: antibody-dependent cellular cytotoxicity, where natural killer cells are recruited to destroy the antibody-coated cancer cell, and complement-dependent cytotoxicity, where the complement cascade punches holes in the cancer cell membrane.²²

Two large phase 3 trials established zolbetuximab’s clinical benefit. SPOTLIGHT tested zolbetuximab plus mFOLFOX6 chemotherapy against chemotherapy alone, and the final analysis showed a statistically significant and clinically meaningful improvement in both progression-free survival and overall survival, with no new safety signals beyond what was already known.²³ GLOW reached a similar conclusion with a different chemotherapy backbone (capecitabine and oxaliplatin), confirming that the benefit was driven by the anti-CLDN18.2 antibody rather than a specific chemo pairing.²⁴ Both trials restricted enrollment to patients whose tumors were CLDN18.2-positive and HER2-negative, reflecting the biomarker-driven approach that is now standard.

Managing Zolbetuximab’s Side Effects

The most common side effects of zolbetuximab are gastrointestinal, which makes biological sense: the drug targets a protein that is also present in normal gastric epithelium. In early cycles, nausea affects the majority of patients. One real-world safety initiative reported that 81% of patients experienced nausea during the first cycle, with about 19% experiencing vomiting despite proactive antiemetic support.²⁵ Another clinical experience initiative found nausea in 54% and vomiting in 25% during cycle 1, with about 63% experiencing some infusion-associated adverse event.²⁶

The encouraging pattern is that these effects are heavily front-loaded. After the first cycle, infusion interruptions dropped substantially (to about 11% in one report) with proactive antiemetic protocols and slower infusion rates.²⁶ Hypoalbuminemia (low blood albumin) is another notable side effect, seen in over half of first-line patients in one cohort. It is thought to result from zolbetuximab-induced inflammation in the stomach lining, causing protein loss through the gut. Patients who had previously undergone gastrectomy experienced less nausea and vomiting, consistent with the stomach being the primary site where the drug causes local irritation.²⁶

Beyond Zolbetuximab: Next-Generation CLDN18.2 Therapies

Zolbetuximab was first to market, but the pipeline of CLDN18.2-directed treatments is expanding rapidly in several distinct directions. Antibody-drug conjugates (ADCs) attach a potent cell-killing chemical to an anti-CLDN18.2 antibody, so the drug is delivered directly to the tumor cell when the antibody binds. IBI343 is one such ADC, carrying the chemotherapy payload exatecan linked to a humanized anti-CLDN18.2 antibody.²⁷ Because ADCs deliver their cytotoxic payload even to cells with relatively low target expression, they may extend the treatable population to patients whose CLDN18.2 levels fall below the strict 75% threshold required for zolbetuximab.

Bispecific antibodies represent another approach. These engineered molecules bind two different targets simultaneously. One group constructed a bispecific antibody that binds both HER2 and CLDN18.2, designed to capture the roughly 18% of HER2-positive tumors that also express CLDN18.2. In tumor spheroid models of gastric cancer, this bispecific outperformed either single-target antibody alone and showed enhanced immune-cell killing activity.²⁸

Perhaps the most ambitious approach is CAR T-cell therapy, which re-engineers a patient’s own immune cells to hunt down CLDN18.2-positive cancer cells. Satricabtagene autoleucel (satri-cel) is the most advanced CLDN18.2-directed CAR-T product. In a phase 1 trial of 98 patients with CLDN18.2-positive gastrointestinal cancers who had already been through prior treatment, the overall response rate was about 39%, with a disease control rate exceeding 90%.²⁹ A subsequent phase 2 trial specifically in gastric and gastroesophageal junction cancer compared satri-cel to physician’s choice of chemotherapy as a third-line treatment. Median progression-free survival was 3.25 months with satri-cel versus 1.77 months with standard chemotherapy, a statistically significant difference.³⁰ These numbers are modest in absolute terms but meaningful for a heavily pretreated population where options are limited. The safety profile was described as manageable, though CAR-T therapies carry their own distinct risks including cytokine release syndrome.

Why Expression Heterogeneity Complicates All of These Strategies

Every CLDN18.2-directed therapy faces the same underlying challenge: the protein is not expressed uniformly across all tumor cells, and expression can shift between the primary tumor and its metastases or change over time with treatment. Abnormal staining patterns, including nuclear and cytoplasmic staining rather than the clean membrane pattern that antibodies need to bind, have been observed in roughly a quarter of primary tumors.⁷ These non-membranous patterns are not targetable by current antibody-based drugs, even though the protein is technically present.

The discordance between primary and metastatic sites, at around 20-25% in published cohorts, adds a practical dilemma. A patient classified as CLDN18.2-positive based on the primary tumor biopsy may harbor metastases with little to no expression. The fact that liver metastases in particular show much lower CLDN18.2 positivity than peritoneal metastases suggests the biology of the organ site influences expression.¹⁶ For patients with liver-dominant disease, CLDN18.2-targeted therapy may be less effective even if the original biopsy looked favorable. Some investigators have advocated for re-biopsy of metastatic lesions before starting targeted treatment, though this is not yet standard practice everywhere. Liquid biopsy approaches that could assess CLDN18.2 status from a blood draw are an active area of research but have not yet reached clinical utility for this specific protein.

These heterogeneity patterns underscore why the field is pursuing multiple therapeutic modalities in parallel. A monoclonal antibody like zolbetuximab needs high, uniform surface expression. An ADC can deliver its cytotoxic payload even with lower expression because only a few antibody molecules need to bind to internalize a lethal dose of drug. CAR-T cells can actively seek and destroy even scattered CLDN18.2-positive cells, though they may miss clones that have completely lost the target. No single approach perfectly addresses every pattern of heterogeneity, and the long-term hope is that combination or sequential strategies might cover more ground than any one agent alone.