Spatial expression of claudin 18.2 in matched primaries and metastases of tubo-ovarian carcinoma of all subtypes

All cases of TOC diagnosed and treated at the University Hospital Erlangen (UKER) between 2003 and 2021 included in the Bavarian Clinical Cancer Registry were screened for their suitability for the study. All patients underwent surgical resection of primary and/or secondary lesions. Availability of tissue in domo, sufficient tumor amount and distinct malignancy were taken as inclusion criteria for our tissue microarray (TMA) cohort. Cases with borderline tumors of epithelial origin were excluded. The formalin-fixed, paraffin-embedded (FFPE) TOC tissue was selected on the basis of medical records and retrieved from the archives of the Institute of Pathology UKER. Fresh hematoxylin and eosin (H&E) slides were fabricated from each tumor block. Taking the initial pathologic diagnosis into account, every case was reevaluated by a board-certified pathologist (R.E.). Histological subtypes were categorized according to the 5th edition of the WHO Classification of Tumors [21]. FIGO and TNM staging was performed according to the current versions of UICC TNM and FIGO classification for the entire cohort [22, 23]. Cases prior to 2014 staged with previous classifications were reclassified within their main category where applicable [24, 25]. Uncertain diagnoses (e.g., high-grade cancer, origin not assessable, or ovarian cancer, not otherwise specified) were additionally confirmed or excluded by immunohistochemistry. The final cohort consisted of 629 patients in total, containing all histological subtypes of TOC. If available, both cancer tissue from the ovary as well as cancer spread from, e.g., the peritoneum, lymph nodes or infiltrated organs, was collected. For the latter group, the term metastases will be frequently employed in the following, which includes both distant metastases (M1; e.g., liver metastasis) and cancer infiltrates disconnected from the main tumor mass in the ovaries (i.e., higher T stages) [23]. Eventually, sufficient FFPE tissue of n = 536 TOCs, n = 385 matching tubo-ovarian carcinoma metastases (TOCM) and n = 93 metastases without primary were included. After processing, cutting and staining the TMAs, n = 529 primary and n = 463 cancer spread samples were evaluable. For n = 619 patients, at least one tumor core was assessable. Figure 1 depicts the workflow from the cohort assembly until the final IHC assessment.

After the final cohort was gathered, randomization in terms of histologic subtype and year of first diagnosis was performed to prevent analytical biases. In the resulting order, the cases were ascendingly allocated to the TMA slots. Next, the areas of interest were marked on the H&E slides. Correspondent punch cores of 1.5 mm in diameter were taken manually from the donor paraffin blocks and transferred to the TMA recipient blocks. Primaries and metastases were split into separate TMAs. To investigate the intratumoral heterogeneity of CLDN18.2 expression, separate TMAs for both the tumor center and periphery were constructed.

Immunohistochemical staining was conducted on 1–2-µm thick sections of formalin-fixed paraffin embedded (FFPE) TMA tumor blocks on the Ventana BenchMark Ultra automated system (Ventana Medical Systems, Inc., Tucson, AZ, USA) after the institute’s standards and the manufacturers recommended protocol. For the CLDN18.2 antibody, a primary anti-human murine monoclonal antibody (clone: 43-14A; host: mouse, isotype: IgG1; Ventana Medical Systems, Inc., Tucson, AZ, USA) was used. Deparaffinization, pretreatment and antigen unmasking with CC1 at 100 °C for 64 min were followed by antibody incubation at 37 °C for 32 min. The binding of the antibody to CLDN18.2 was visualized with the optiView DAB IHC Detection Kit (Ventana). Eventually, subsequent counterstaining with hematoxylin and Bluing Reagent (Ventana) was conducted. For on-slide positive controls and the establishment of the antibody, human gastric mucosal tissue was employed.

After review of morphology in the TMA H&Es IHC, assessment was performed by an experienced gynecopathologist (R.E.) and a resident trained in gynecopathology (P.W.) blinded to the clinical data. Decision making was consensus based and on the basis of the phase III SPOTLIGHT biomarker study criteria [15]. The assessment was done manually with a Zeiss Axio microscope (magnification of × 50, × 100, × 200, and × 400). The percentage of positive tumor cells and the staining intensity were evaluated with regard to only the cell membrane. The percentage of stained tumor was visually estimated in 5% increments. Staining intensity was measured semiquantitatively and divided into four categories (0 as no staining, 1 as weak, 2 as moderate, and 3 as strong). CLDN18.2 positivity was defined as biomarker expression of ≥ 75% in tumor cells with moderate-to-strong membranous staining [15]. To quantify the whole staining pattern, the IRS score after Remmele and Stegner, taking staining percentage and intensity into account, was also calculated [26]. A minimum of 100 viable tumor cells was required for assessment.

Statistical analysis was performed with GraphPad PRISM version 8 (GraphPad Software, Inc., San Diego, CA, USA). Paired t tests, Wilcoxon tests and Mann‒Whitney U tests were used for the comparison of intra- and intertumoral matched pairs and expansile vs. infiltrative subtypes, respectively. Varying expression patterns between subtypes were analyzed with the Kruskal‒Wallis test. A p value < 0.05 was considered significant.

The histological subtype distribution in the final cohort was as follows: high-grade serous carcinoma n = 445, low-grade serous carcinoma n = 40, endometrioid carcinoma n = 43, mucinous carcinoma n = 43, clear-cell carcinoma n = 36, carcinosarcoma n = 17, malignant Brenner tumor n = 4, and undifferentiated, not otherwise specified n = 1. Table 1 gives clinicopathological data about the cohort. Of the total of n = 478 metastases, n = 384 were located in the peritoneal cavity, n = 17 were regional lymph node metastases and n = 58 originated from the parenchyma of pelvic or intraabdominal organs. N = 37 metastatic samples were from relapses. According to the 8th edition of the UICC TNM classification for TOC, n = 19 were distant metastases [23]. More detailed information about the numbers, localizations and size of the metastatic tissue are provided in the Supplementary Table S1.

From overall n = 619 evaluable patients, a total of n = 44 cases (7.1%) showed membranous CLDN 18.2 staining to some extent. The vast majority of cases (n = 575) was entirely CLDN18.2 negative. Figure 2 illustrates the typical appearance of all nonmucinous TOC subtypes in CLDN18.2 immunohistochemistry. Applying the SPOTLIGHT III criteria, CLDN 18.2 was positive in n = 28 cases (4.5%) in at least one of the tumor sites [15]. In regard to primaries, CLDN18.2 was detected in 4.1% (21/515) of TOC centers and in 3.6% (18/498) of TOC peripheries. In metastases, the positivity quota was only 0.9% (4/453) in the center and 0.7% (3/405) in the periphery. Apart from 1 out of 445 high-grade serous carcinomas and 2 out of 43 endometrioid carcinomas, CLDN18.2 was highly significantly restricted to MTOCs (Fig. 3).

Among MTOCs, however, CLDN18.2 positivity rates were 45% (18/40) in the tumor centers and 36.6% (15/41) in their peripheries. The mean IRS score in primaries was 6.125 (SD 4.57; 95%-CI: 4.67–7.59) in the center and 6.81 (SD 4.41; 95%-CI: 5.41–8.2) in the periphery. Paired t test did not reveal a significant difference in IRS scores between the two sites (p value = 0.1881). Overall intratumoral concordance, meaning that the matched pairs of center and periphery were both negative or positive with respect to the SPOTLIGHT III criteria, was 76.9% (30/39) (Supplementary Figure S1). Figure 4 displays the central and peripheral CLDN 18.2 expression of selected MTOC cases.

Positivity rates for the corresponding mucinous tubo-ovarian carcinoma metastases (MTOCMs) were 31% (4/13, center) and 23% (3/13, periphery). In the central area, the mean IRS score was 5.54 (SD 4.74; 95% CI: 2.68–8.4) compared to 5.08 (SD: 3.59; 95% CI: 2.91–7.25) in the peripheral area. As in primaries, the IRS score did not differ significantly between the localizations (p value = 0.6563). 92% (12/13) of the MTOCM pairs were intratumorally concordant (Supplementary Figure S2).

A total of n = 12 pairs of primaries and corresponding metastases of MTOC could be matched. The couples were considered concordant in their CLDN18.2 expression if positivity was granted for both localizations in at least one of the tumor areas (center or periphery). Thus, 58% (7/12) of the couples were concordant, whereas 42% (5/12) were discordant by definition. However, the mean IRS scores in the center and periphery of primaries and metastases did not vary significantly (p value (center) = 0.3594, p value (periphery) = 0.1875) (Supplementary Figure S3). Figure 5 shows examples of concordant and discordant combinations of primary and metastatic lesions.

The current WHO classification 5th edition no longer divides mucinous carcinomas into three grades but distinguishes between an expansile and infiltrative growth pattern, which has been shown to better correlate with prognosis [27]. Expansile growth refers to extensive back-to-back gland proliferation with preserved differentiation and is associated with lower stage, reduced invasiveness and better prognosis [28, 29]. Of the 43 cases in our cohort, 31 showed an expansile growth pattern, and 12 showed an infiltrative growth pattern. We observed CLDN 18.2 expression to be emphasized in the expansile subtype, which shows a high similarity to gastrointestinal epithelium. The expression then tends to decrease during the process of dedifferentiation toward infiltrative growth but is nonetheless preserved to a noteworthy extent in this type. This is reflected by a positivity rate of 58% (18/31) in expansile and 42% (5/12) in infiltrative mucinous ovarian carcinomas. Their mean IRS scores did not differ significantly (p value (center) = 0.2842, p value (periphery) = 0.1085). Supplementary Figures S4 and S5 provide case examples of the staining behavior of the two subtypes in both primaries and metastases.

In summary, we were able to robustly support the previous scientific findings. Our results are consistent with the few preexisting data that have linked CLDN18.2 expression exclusively to the mucinous subtype of TOC. The overall CLDN18.2 positivity of 4.5% (n = 28/619), allows the studies’ first conclusion that CLDN18.2 will supposedly not display a general target for personalized therapy in patients with TOC to be drawn quickly. In the direct comparison, CLDN18.2 was more frequent in primaries (n = 24/529, 4.5%) than in metastases (n = 4/463, 0.9%), which could supposedly be explained by a comparably higher percentage of high-grade serous carcinoma among the metastases.

Simultaneously, this stringent restriction to the mucinous subtype offers opportunities from a diagnostic point of view. For example, the differential diagnosis between endometrioid and mucinous carcinomas can be challenging, since both subtypes can mimic the morphologic features of one another. On the other hand, we noticed a slight emphasis of CLDN18.2 in gastrointestinal differentiated MTOCs. Although the classification of mucinous borderline tumors in intestinal (IBMT) and endocervical types (EBMT) has in the meantime been abandoned, a study in 2013 reported a similar pattern for borderline tumors. They postulated a Mullerian origin of EBMT, which explains the absence of CLDN18.2 as a marker of gastric lineage [30]. Our findings point in the same direction, leaving behind questions about the exact definition and origin of mucinous carcinomas, which are still subject to scientific debate [31, 32]. Another unsolved problem in this context has always been the difficult distinction between MTOCs and metastatic lesions from the gastrointestinal tract. Recalling the results of Wang et al., CLDN18.2 can help differentiate metastases from the upper gastrointestinal tract (CLDN18.2 +) from those of the lower gastrointestinal tract (CLDN18.2 −) [20]. Unfortunately, it cannot cover the demand for a much-needed marker to reliably distinguish MTOC from metastatic mucinous carcinomas, a decision which, in case of doubt, has to be made on the basis of clinical and imaging criteria.

The mucinous subtype is exceptionally rare, accounting for only 3% of ovarian carcinomas [33]. At the same time, MTOCs differ distinctly from other subtypes in their biological behavior. Compared to the dominant high-grade serous carcinoma, they tend to be diagnosed at an earlier stage, while the affected women are usually younger, with one-quarter of the patients being under 44 years of age [34]. Meanwhile, mucinous carcinoma is classically characterized by a bad response to platinum-based chemotherapy, leading to a very poor outcome in the case of advanced disease [35]. The two previous studies performing CLDN18.2 IHC in TOC raised with 22 or 17 cases about half of the patients, respectively [8, 20]. In addition, their study objectives differed from ours, and only mucinous carcinomas were included. Wang et al. reported a positivity rate of 86.4% in TOC, which at first glance differs greatly from our observations. However, they defined a cutoff value of 50% of the tumor cells stained with moderate intensity to assign cases positive. Transferring this scoring system to our cohort would change the positive percentage from 46 to 56%. Under the condition of moderate staining (2 +) in at least 60% of the cells, Sahin et al. reported a positivity rate of 24%. As their findings widely spread from one another, our study contributes additional information to estimate the actual CLDN18.2 frequency in MTOC. Since data from the phase II study of Tureci et al. suggest a superior efficacy of Zolbetuximab in patients with CLDN18.2 expression in ≥ 75% of the tumor mass, a rather high cutoff value seems appropriate [36].

Overall, we were unable to find a strong emphasis of CLDN 18.2 expression at a particular tumor site. As the tumor periphery and metastases are thought to be the relevant scenes of invasion, the question about CLDN18.2’s role in tumor progression and dissemination remains unanswered. Interestingly, claudin subtypes 3 and 4 have been particularly linked to the invasive character of TOC [37]. A gene analysis found them to be among the most upregulated genes in TOC cells [38]. Knockdown of CLDN3 or CLDN4 in tumor cell lines results in reduced tight junction formation and increased invasiveness [39]. Several immunohistochemical studies could associate high expression levels of CLDN3 and CLDN4 with poorer survival [40‐42]. These findings support the hypothesis that claudin-mediated alterations in tight junction integrity with subsequent loss of polarity and cellular adhesion represent hallmarks of carcinogenesis [43].

The recently published results of the SPOTLIGHT and GLOW studies promote large expectations [15, 16]. In a large set of locally advanced or metastasized gastric cancers, a regimen of Zolbetuximab + mFOLFOX6 (5-fluoruracil plus oxaliplatin and leucovorin) showed significant superiority compared to the placebo-controlled group. The risk of progression or death was reduced by 25%, and the median PFS was prolonged by approximately 2 months. Median overall survival rose from 15.54 to 18.23 months, resulting in a hazard ratio concerning the risk of death of 0.75 [15]. These outcomes raise the hope of potentially being transferable to other CLDN18.2-positive entities, such as MTOCs. Its pronounced expression in the predominantly gastrointestinal differentiated MTOCs illustrates the close biological relationship of the two entities that might come along with a similar vulnerability toward targeted therapy. As our study provides evidence for the occurrence of CLDN18.2 at all tumor sites, a therapy with, e.g., Zolbetuximab might be able to temper the current therapeutic deficit in advanced MTOC. However, in view of the epidemiology of the disease, realistic expectations should be maintained. MTOCs, which are very rare anyway, are usually diagnosed at an early stage and can then be adequately treated surgically. Finding a sufficient number of patients for clinical trials has already proven to be complicated and the potential target group for an anti-CLDN18.2 therapy may be too small to arouse the serious interest of trial sponsors [44].

The restriction to immunohistochemical analyses only is obviously a limitation of this study. For a deeper understanding of CLDN18.2 gene expression in TOC, concomitant mRNA analyses would be desirable. On the other hand, with high availability and rather low cost, IHC as a diagnostic means in practical use has several advantages. Furthermore, it proves the expression of markers on a protein level. Beyond that, companion diagnostic tests for therapeutic antibodies, which in the case of CLDN18.2 already exist, are usually based on IHC [45]. Another point of criticism to mention is the relatively small statistical power of our analysis. Despite the long retrospective timeframe, only 43 cases of MTOC could eventually be assessed. Since mucinous carcinomas, compared to other subtypes of TOC, have a rather small tendency to spread metastatically, the metastases also did not reach a statistically robust number. Nonetheless, the approach succeeded in revealing certain tendencies in expression patterns and contributing additional insights to the yet scarce information on the topic.