Dr. Graham: The Claudin proteins are a family of at least 27 transmembrane proteins. They're composed of four transmembrane helix loops — four helix domains as the protein traverses the extracellular membrane — and they are a major component of tight junctions in cells. Consequently, they are expressed throughout the body and display organ-specific or tissue-specific expression patterns.

Claudin 18.1 and 18.2 are two isoforms of Claudin 18. They differ in their amino acids, with eight of 51 amino acids showing a difference. They're very similar — Claudin 18.2 is expressed in the stomach and 18.1 in the lung. That bears some relevance when you're thinking about off-target effects. Across the family of Claudins as a whole, though, they're expressed all throughout the body.

Claudin 18.2 is present both in primary gastric adenocarcinomas and in metastases, making it an attractive target for therapeutic intervention. Accumulating data have shown that accurate Claudin 18.2 assessment can be performed both on the primary tumor and on the metastasis — we'll touch on this in some subsequent slides. Because the protein is located at the level of the membrane, one would anticipate membranous staining to correspond to a positive result. And that's exactly what we see.

Prevalence

Dr. Graham: Looking at the global prevalence of Claudin 18.2 across two Phase III trials, Claudin 18.2 expression is seen in approximately 40% of gastric cancer cases at the level of expression that's considered positive — that's 75% or greater. There are some smaller percentages of cases that show a lower level of expression, and about a quarter of cases are not expressing Claudin 18.2 at all.

When we look at positivity across tumor location — whether it's a proximal or distal gastric cancer — the prevalence of expression is about the same. Similar prevalence whether you're in the stomach or the GE junction. And the prevalence doesn't vary significantly across the types of gastric cancer using the Lauren classification. So this is a marker that is positive in a substantial proportion of gastric cancers and shows a similar level of positivity regardless of location or classification system.

When you look at biopsies and resections from these two Phase III trials, there's almost equal prevalence of Claudin 18.2 expression. And whether you sample the primary or the metastasis, there's approximately equal expression.

Claudin 18.2 Among Gastric Cancer Biomarkers

Dr. Graham: We're going to turn our attention to comparing Claudin 18.2 to some other biomarkers. The commonly assessed biomarkers in gastric cancer include Claudin 18.2, HER2, PD-L1, and MMR. Many of us will be familiar with MMR, especially if you're evaluating colon cancer cases — you'll see nuclear expression of MMR proteins. PD-L1 is everyone's favorite, and really turned our lives upside down a few years ago. We won't elaborate very much on that, but recognizing that it's a biomarker with a complex history and can be challenging to interpret.

While there are challenges with Claudin 18.2 too, I think these are easier to interpret. PD-L1 calls for membranous expression on immunohistochemistry. HER2 is generally scored using the GE junction or gastric criteria, first established in the ToGA trial. And Claudin 18.2 is assessed by looking at staining intensity — 2+ or greater — and the percentage of cells. I think it's more intuitive to interpret; certainly easier than PD-L1.

Claudin 18.2 Targeted Therapies

Dr. Graham: There are a number of Claudin 18.2 targeted therapies that are approved or under investigation. This is an area of active interest, which probably corresponds to the prevalence of Claudin 18.2 in gastric cancer. This is a target that is worthwhile knowing about — and good on you for taking the time to invest in learning more.

Available Antibodies & Assays

Dr. Graham: A number of different antibodies have been developed to identify Claudin 18.2 expression in tissue. The first two come from Ventana/Roche, with several other companies' options also available. Most of these are monoclonals, but there are polyclonal options as well.

The FDA-approved companion diagnostic is the Claudin 1843-14A clone from Roche, which was tested in the two clinical trials. When you look at its analytical performance — across different lots, instruments, and days of testing, really putting it through its rigor — concordance rates were approaching 99%. Clinical performance showed acceptability rates around 95%. This is the assay we use in our clinical practice to identify Claudin 18.2-positive patients.

In the global RING study, which looked at multiple antibodies across multiple platforms, reproducibility was assessed. The Ventana Claudin assay showed quite high analytical performance, and there was equivalent performance with some other antibodies — LSBio and Novus, depending on the platform. An important finding in that study was that the high quality of performance among pathologists was linked to a strong emphasis on training, which is exactly what we're doing here today.

Along these lines, it's useful to look at the Nordic QC data. They ran extensive comparisons looking at Claudin 18.2 across a large number of labs and found an 87% pass rate. This gives us a window into what happens in the real world — despite excellent analytical performance in controlled environments, there was more variability in acceptability in practice. LDTs with concentrated antibodies actually performed poorly, with a 46% pass rate. So if you're going to use anything other than the companion diagnostic, attention to the details in validation will be really crucial to ensure reliable results.

Interpretation Principles

Dr. Graham: The antibody, I think, is very straightforward to interpret. At zero staining or no staining, the background is very clean. Weak staining is scored as 1+, and then you see moderate and strong staining. An on-slide positive control is routinely used to support interpretation.

What we're looking for: all invasive tumor cells form our baseline — the denominator. The numerator is those invasive tumor cells that are positive with membranous expression of greater than moderate. That membranous expression may be apical, circumferential (partial or complete), basolateral, or lateral. The positive threshold for this indication is 75%, and you need a minimum of at least 50 tumor cells for the test to be valid.

Normal gastric mucosa is really helpful to look at — it gives you strong 3+ positivity and serves as a great internal control. You can recognize this at low power and then move from there to evaluate the target tissue on the slide.

Interpretation: Practical Considerations

Dr. Graham: The important things to consider as you go through interpretation: pay attention to the denominator — approximately how many tumor cells are present on the tissue sample. Tumor cells are characteristically smaller and haphazardly arranged, which presents a layer of difficulty. Pay careful attention to the H&E section to assess how much tumor there is.

A few other tips to be aware of: signet ring cell carcinoma can be a challenge even on H&E, so pay attention to how many cells there are and where they are when assessing your denominator. Inflammation can be tricky, especially with signet ring cells — if you have lots of plasma cells, you want to be really clear about what you're looking at. Using IHC for epithelial markers is a useful approach if you need help clarifying the denominator. I haven't actually had to do that in practice, but conceptually it's a good fallback.

The inclusion criteria: we're counting membranous staining only — complete or partial — and this includes basal, basolateral, and apical staining. What we're not counting: non-neoplastic glands, necrosis or necrotic tumor, mucin, or any cytoplasmic staining.

You'll find some areas where staining is weak or moderate, or where tumor has infiltrated through the muscularis propria. Paying attention to where those cells are located is helpful. Areas with weak or moderate staining must be included in the denominator (total tumor cells), but may not meet the threshold for the numerator. You can also encounter areas where necrosis or mucin creates what almost appears to be a smudging artifact at low power, or tissue folds that add difficulty. Only count viable tumor cells — don't include mucin, necrotic debris, or cytoplasmic staining.

Tumor Heterogeneity

Dr. Graham: When you're looking at cases with a range of staining intensity, you can have areas that are completely negative, areas that are strongly positive, and staining in between. This does happen in real cases. I haven't found it particularly challenging, but the key is just to be aware that it does happen.

When considering tumor heterogeneity, there are four tips: first, be aware of spatial heterogeneity — you may see differences in staining between the mucosa and the muscularis propria, for instance, or variation in intensity. Second, if your estimated percentage is close to the cutoff, seek consensus with peers. I use a threshold of roughly 65% to 85% — if I'm in that range, I'll ask a colleague. If we're both on the same side of positive, I go with that result. If we're on opposite sides, we find a tiebreaker.

Some excellent data from colleagues in Padua, looking at intratumoral heterogeneity, found that with increasing numbers of biopsies, sensitivity and specificity for identifying the 50% cutoff both improved. This underlines what we tell our endoscopists: if you take multiple biopsy fragments, we're more likely to get a reproducible, reliable result. With fewer biopsies, the chance that heterogeneity has skewed the result remains.

Key Interpretation Takeaways

Dr. Graham: Key takeaways when interpreting Claudin 18.2: you're looking for membranous expression, a percentage of positive tumor cells, at least moderate intensity or greater. In my practice, and I think this is best practice, report both the percentage of cells and a positive or negative result. This way, if the indication changes or there's emerging data, you don't have to go back and review cases again to look for where they fell — whether 50% or 65%. Putting the raw percentage in the report is a helpful practice. 

PRE-ASSESSMENT: Interactive Case Review

Dr. Graham: We're going to go through a pre-assessment discussion so you'll get an opportunity to see some cases and start to warm up on interpretation.

Case 1: You can see on the H&E section a tumor with some surface-level necrosis, probably at the level of the GE junction. Lots of infiltrating tumor. On the corresponding Claudin image — ignoring that surface area where there's fibrin and debris — there's extensive expression of Claudin 18.2. We want to get a little closer to ensure we're looking at membranous expression across those tumor cells, but the vast majority of the tissue appears positive.

[Audience enters responses via polling system]

The correct answer is 96%.

Case 2: Gastric mucosa on the surface, with the submucosa filled with tumor, extending more deeply into the muscularis mucosa. Looking closer, you can see lots of tumor forming abortive anastomosing glands with Claudin 18.2 expression shown in brown. My approach at low power: I ask myself, if this were all tumor at 100%, are approximately half the cells staining? If yes, I go in closer to fine-tune. If not half, is it a quarter? That gives me a useful binning approach. For this case, the answer is 35% — a negative result for this indication.

Case 3: A rectangular section where I apply the same approach. Are virtually all cells staining? No. About half? No. I work toward a quarter, and that helps me bin into the right range. Even if I'm off by 5% to 10%, I'm at least directionally close. This one has an almost plasmacytoid morphology, with expression around these tumor cells. The answer is 5%.

TRAINING SET: Post-Assessment Discussion

Dr. Graham: We're going to do this the old-fashioned way — raise your hand for A, B, and so on.

Case 1: At low power, tumor with a very tubular-papillary configuration. On higher power, well-developed papillae with thin fibrovascular cores surfaced by hypochromatic tumor cells. The Claudin 18.2 expression at low power looks essentially positive across almost all tumor cells — clearly more than 50%, predominantly moderate or greater. There's one area that's weaker, but the lion's share is positive. This is reassuring. The correct answer is 85% — positive.

Case 2: Tumor concentrated at the surface, but more tumor is interspersed throughout the underlying stroma. At low power, expression is mainly at the surface — I'm already thinking this will be less than 50% of tumor cells positive. Closer up, you can see tumor — not just at surface level — and the expression is principally at the surface. About a quarter of the tumor cells, I'd say. Some areas show signet ring type morphology with nothing to weak staining, possibly a couple of moderate cells but clearly a minority. I'm pretty convinced this is negative. The answer is 35% — negative.

Case 3: At low power, tumor cells interspersed with inflammation.

[Audience raises hands — virtually everyone votes for 0%]

Correct. The answer is 0%.

Case 4: Two tissue fragments — one more vertical, one horizontal. There's staining in one fragment but much less in the other. At higher power, tumor cells clustered around blood vessels, some edematous areas, some positive cells at moderate or greater intensity, with intervening areas of tumor cells. Integrating both fragments: one showing more expression, one much less. The key question — is it half or less? I think slightly less than half. Then this field, a lot more staining; this field, some staining, some not; this field, tumor cells are negative. Taking it all together:

[Audience votes: 10% — 1; 25% — several; 40% — several; 65% — zero]

The correct answer is 40%. This is the most challenging of the five cases. Crucially, everyone recognized it would be negative — the split in the room shows where everyone fell on the exact percentage, which is exactly where you'd expect variability to occur on a challenging case.

Closing Remarks

Dr. Graham: I think you'll see that you can do this with confidence. Your answers, where there were differences, were usually within 10% to 15% of each other. This reinforces an important point: if you think you're at about 65% or about 60%, that's worth showing to colleagues. If you think you're at about 85%, that's also worth showing. Over time, you'll build up confidence and consensus around whether a case is positive or not.

The Claudin 18.2 test has validated staining characteristics. Gastric epithelium is a terrific control — widely available, so you can use it in your lab. If you have intestinal metaplasia or intestinal epithelium, that can serve as a control for moderate staining. In my experience, and as you've seen across almost 12 examples here, the staining is really clean — not a particularly high background.

There's a website you can go to — the link is included in the slides — to look at other examples and to see more information about the immunostain. Your feedback is also really important to the organizers. Take the time to fill out the paper survey; it'll be picked up from the desk. And if there are residual questions, we're going to take those immediately after.

I will try to get as many questions as possible. Thanks for attending.

AUDIENCE Q&A

Q: Are you doing Claudin, PD-L1, HER2, and MSI on all gastric cancers?

Dr. Graham: That's correct. In Rochester, we're doing it on all cases. A natural follow-up to that question might be what our prioritized order is — which you're about to ask me. Our prioritized sequence begins with HER2 and PD-L1, followed by Claudin and MMR. That's our prioritized sequence, but we do them all automatically. We face the block once and are usually able to get them all complete on biopsy.

Q: Do you also score the dysplasia?

Dr. Graham: Great question — the answer is no, only the invasive carcinoma.

Follow-up question: You said the minimum number of cells needed is 50. If you have a small cluster of highly positive cells but fewer than 50, would you just exclude that case?

Dr. Graham: That's a great question. I haven't encountered it in practice, but what I typically do in this scenario — and this comes up with some other biomarkers — is I will report the sample as insufficient for analysis, but include a comment that while there are fewer than 50 cells, all of them are positive. Usually what happens next is they biopsy again.

Follow-up: Has this IHC been validated for cytology samples?

Dr. Graham: To my knowledge, no.  

Q: If a biopsy is positive for Claudin 18.2, do you go back and repeat the stain on the resection specimen?

Dr. Graham: In our practice, we don't. If we get a positive result, we consider that patient to be positive. What can be interesting is if the reverse happens — the biopsy was negative, and then at an interval there's a metastasis. In our practice, the oncologists typically will order the stain in that setting, but it won't be automatic — they'll need to request it.

Follow-up: With spatial and tumor heterogeneity in mind, do you pick one of the best slides, or do you stain all slides of the invasive carcinoma?

Dr. Graham: If there are multiple blocks, we pick one — the most representative block. What we think represents the cellularity and bulk of the tumor. Recognizing that this introduces some bias based on what the pathologist thinks. But that has been the approach we've used for all biomarkers all the time.

Q: Do you get pushback from insurers? Do you get pushback from institutions not wanting to do these stains because it's yet another one? We get fights for PD-L1 unless criteria are very stringent.

Dr. Graham: It's a really important comment. I think there will probably be institutional and regional variation. In our practice, we have not had pushback. What we've done to support our practice is, when we decide to adopt a biomarker, we have a meeting with our medical oncology team and put to paper in writing what drives the order and how we'll use it. That drives our policy framework. If an issue comes up for a patient, the oncologist then does their work with the insurance company. I would say we have not had that issue with almost any of our digestive system biomarkers. Hopefully that's helpful.

Q: I want to take this back to my lab. My lab manager is going to ask whether Claudin 18 expression is common enough to support validation. What are my responses to help get this adopted?

Dr. Graham: Thanks for that question about validation. The frequency of positive results is actually pretty high — positive results will occur in about 4 of every 10 cases. So if you're getting a substantial volume of gastric cancer, you won't have difficulty finding positive cases for validation. Positive controls are also common — if you get lots of sleeve gastrectomies, those will be a ready source. And negatives are almost never a problem to find.

Using the CAP framework, you're going to need 20 positives and 20 negatives. This is something many labs will be able to do if they have substantial gastric cancer volume. Another thing to discuss with your lab manager: HER2 and how you'll manage the workflow with potentially small biopsies. It's suboptimal to face the block for one stain and then come back two days later for another. Talk with your oncologist about what the prioritized sequence will be, and how you'll manage that from a workflow perspective. Make sure they have awareness of the implications beyond just the validation itself.

Q: I presume you see a lot of consults at Mayo with outside pathology. Do you ever repeat the Claudin stain done on the outside?

Dr. Graham: That's a great question. I almost never repeat external biomarkers if there's an on-slide control that appeared to work. But if there isn't an on-slide control and it's negative from the outside, yes, I would probably repeat that in that setting.

We do sometimes get people sending us cases just for the biomarker alone — that's part of our practice. And I would say once the approval happened, we actually had a lot of Claudin cases coming in. In the first four months, we might have had 500.

Q: Do you see more enrichment of positivity in signet ring versus regular adenocarcinoma? I would expect that you would.

Dr. Graham: The initial data show no difference in prevalence based on histologic subtype. That said, the data are still emerging. For now, histologic subtype in the context of gastric cancer doesn't appear to affect the prevalence of a positive result.

[Audience member: I had an abstract on Claudin-RGAP26 fusion that's enriched in a subset of younger patients, mostly female. I would expect more in signet ring.]

Dr. Graham: Yeah, well, it could be — but it's just not played out in the data yet. One interesting thing here is that we're looking at Claudin 18.2 expression at the membrane level. Of the various drivers of gastric cancer, you can have Claudin 18.2 expression across various molecular mechanisms. Claudin fusions can be seen, but that's not the only mechanism. To put it another way — HER2 overexpression does not always correspond to HER2 mutations. We're just looking at the protein phenomenon right now, and for now, that's what the data show.

[Audience member: I thought it was an adhesion protein, that's why I asked. Would you expect Claudin 18.2 expression in other signet ring tumors, like lobular carcinoma of the breast?]

Dr. Graham: It is an adhesion protein. And that's a good question — I think you would have to study it to see.

Q (Dr. Lazenby): When we get our GEJ resections, it's usually post-chemoradiation. Is it still relevant to do Claudin 18.2 there? Does chemoradiation affect the staining artifactually? Do you recommend not doing it on those resections?

Dr. Graham: This is a good question. I'll have to look back at the trial. Angela, do you want to weigh in on what you saw with neoadjuvant?

[Angela, from audience: There's actually a high rate of concordance pre- and post-chemoradiation treatment.]

Dr. Graham: That's perfect. In our practice, we do have a lot of neoadjuvant cases like you do, Dr. Lazenby. And sometimes we get pathologic complete responses. Our bigger challenge is actually having enough tumor left on the resection specimen. But it's good to know the results are concordant based on those data.