Freshly sectioned tumor samples were formalin-fixed and paraffin-embedded (FFPE). Tumor samples were obtained from the biobank of the Institute of General and Molecular Pathology and Pathological Anatomy Tuebingen. The samples used in this study were obtained from 55 patients with mCRC, who were treated at the University Hospital of Tuebingen between 2010 and 2023. At the time of sample acquisition, all patients had reached the metastatic stage (stage IV according to the Union Internationale Contre le Cancer (UICC)), bearing distant metastases either in the liver or lung. The specimens included tissue from primary colorectal tumors (n = 33), liver metastases (n = 12), lung metastases (n = 9), and peritoneal metastases (n = 1).

Additionally, we analyzed sequential samples from a randomly selected subset of n = 7 patients.

All relevant clinical data were extracted from the digital patient records. Data on PFS and OS were available for 47 and 49 patients, respectively, and were included in the corresponding analyses. Patient follow-up was conducted until death or the last documented clinical contact.

To identify suitable patient tissue specimens, hematoxylin and eosin (H&E) stained histological sections were examined microscopically to confirm the presence of tumor tissue, as verified by a board-certified pathologist. The tissue samples included both resection specimens and core needle biopsies. Two tissue cores (if possible) with a diameter of 1.0 mm were then extracted from corresponding FFPE blocks to construct a tissue microarray (TMA). To ensure representativeness, we aimed to sample tumor areas from both the center and the invasive front whenever possible, making the TMA heterogeneous by design. In certain cases, however, only liver biopsies or limited tissue material were available, thus precluding the extraction of two cores. In total, 124 cores including 2 control cores (healthy liver) were placed on a single TMA block.

Immunohistochemical staining was performed on FFPE tissue sections (2.5 µm) mounted on coated TOMO slides. Slides were deparaffinized at 72°C using EZ-Prep solution (Roche, Basel, Switzerland), followed by heat-induced epitope retrieval at 100°C for 64 min in a slightly basic Tris-based buffer (CC1, Ventana Roche). After cooling, slides were incubated with the primary antibody RBT-CD276 (Medac/Bio SB), a rabbit polyclonal antibody specific for B7-H3 (CD276), at 37°C for 32 min. Signal detection was performed using the OptiView DAB Detection Kit (Ventana Roche). This detection system includes incubation with a hapten-based amplification complex consisting of a cocktail of polyclonal secondary antibodies (goat anti-rabbit IgG, goat anti-mouse IgG, and goat anti-rabbit IgM), followed by an HRP-conjugated multimer for 8 min. Visualization was achieved using 3,3′-diaminobenzidine (DAB) chromogen for 8 min. Sections were then counterstained with hematoxylin (20 min), blued (8 min), washed in distilled water, dehydrated through graded alcohols, cleared in xylene, and mounted with Cytoseal (Epredia, Portsmouth, NH).

Stained slides were scanned using a Pannoramic MIDI Scanner (3DHISTECH, Budapest, Hungary), and digital image analysis was performed using the SlideViewer software (3DHISTECH) for optical examination. The staining intensities in both tumor tissue and stroma were assessed by an advanced pathology resident (VG) supervised by a board-certified pathologist (CMS) using an established scale from 0 to 3: 0 (no staining), 1 (weak staining), 2 (moderate staining), and 3 (strong staining) [20]. Two replicate samples of each patient were averaged, and the Histo-Score (H-Score) was calculated using the following formula [21]:

H-Score = (% of unstained tumor cells × 0) + (% of weakly stained tumor cells × 1) + (% of moderately stained tumor cells × 2) + (% of strongly stained tumor cells × 3).

For further analysis, the patients were divided into three equally sized groups based on H-Score levels, with the following ranges:

• Tertile 1: Low H-Score (0–190)

In cases where patients had identical H-Score values at the boundary between two tertiles (190 or 285), these patients were randomly assigned to one of the adjacent groups. This randomization strategy was applied to ensure equal group sizes for subsequent analyses.

Continuous data were assessed for distribution. For testing normality, the Shapiro-Wilk test was used. Group differences were analyzed using ANOVA for normally distributed data and the Kruskal-Wallis test for non-normally distributed data. Post hoc pairwise Dunn tests with Benjamini-Hochberg correction were applied where appropriate. Fisher’s exact test was used to compare categorical variables between groups. Correlations were evaluated using Spearman’s rank correlation.

Survival analyses were performed using the Kaplan-Meier method. OS was defined as the time from diagnosis to death or to the last contact for censored patients. Patients who were alive at the time of analysis were censored, and deceased patients were considered events. Kaplan-Meier curves were generated using the R survival package, and differences between survival curves were assessed using the log-rank test.

PFS was defined as the time from diagnosis to documented disease progression or death. Patients with missing longitudinal data were excluded. Since all patients with available follow-up data experienced a progression event, no censoring was applied. A log-rank test was used to compare PFS between groups.

All statistical tests were conducted at a significance level of p < 0.05.

All statistical analyses and figure generation were conducted in R Studio (R version 4.4.1).

For the assessment of B7-H3 expression, tumor samples from 55 patients were analyzed. The mean age of the patients was 60.1 years. The male-to-female ratio was 1.29:1. At the time of analysis, 85.7% of patients were deceased, with a median OS of 35 months and a median PFS of 12 months. The majority of patients (62.3%) presented with T-stage T3, followed by T4 (28.3%), indicating the depth of tumor invasion. Regarding lymph node involvement, 44.4% had N-stage 1, and 27.8% had N-stage 2. Most patients (78.0%) exhibited a histopathological tumor grade (G) of 2. At the time of the initial biopsy, all patients (100%) had already reached the metastatic stage, and at the time of diagnosis, 87.3% had metastases. A higher proportion of left sided primary tumors (63.0%) was observed compared to right sided tumors (37.0%). The patients underwent a median of 3 lines of chemotherapy, and 47.3% had received chemotherapy prior to the initial biopsy used for our TMA. With respect to genetic and molecular markers, 11.1% of patients exhibited tumors with MSI (MSI-high or MSI-low), 61.4% harbored a KRAS-, 2.9% a NRAS-, 8.8% a BRAF-, 19.1% a PIK3CA- and 80.0% a p53 mutation. HER2 expression scores were available for only 15 patients, of whom 4 showed a score of 2, and 2 showed a score of 1. The mean serum CEA level at diagnosis was 246.4 μg/L, while the mean serum CA19-9 (carbohydrate antigen 19-9) level was 1285.6 kU/L (Table 1).

To demonstrate that the studied cohort represents a typical patient population with CRC, an analysis of established risk factors was conducted, examining their influence on PFS and OS.

A higher T-stage, according to the TNM classification, reflects deeper tumor invasion and is associated with worse outcomes [22]. In our cohort, a Logrank test revealed no significant difference in PFS between T-stage groups (p = 0.22), nor in OS (p = 0.13) (Supplementary Figure S1A).

Nodal metastases (N-stage) are a well-established adverse prognostic factor in colorectal cancer [23]. In our cohort, no significant association between N-stage and PFS was observed (p = 0.11). For OS, however, the difference between groups was significant (p = 0.01) (Supplementary Figure S1B).

The presence of lymphatic invasion (L1 according to the TNM classification) is a known independent adverse prognostic factor in colorectal carcinoma patients with liver metastases [24]. While no such difference was observed for PFS (p = 0.32), OS differed significantly between patients with and without lymphatic invasion (p = 0.02) (Supplementary Figure S1C).

A poor histological grading is known to be an independent prognostic factor in colorectal cancer [25]. In our cohort, PFS differed significantly according to tumor grading (p = 0.01), indicating a relevant impact of tumor differentiation. For OS, a trend toward worse outcomes with poor grading was observed (p = 0.14) (Supplementary Figure S2A).

The presence of metastases at diagnosis is known to be an adverse prognostic factor in colorectal cancer patients [26]. While no significant difference was observed for PFS in our analysis (p = 0.09), patients presenting with metastases at initial diagnosis of CRC had significantly worse OS (p = 0.01) (Supplementary Figure S2B).

Several studies have highlighted the prognostic significance of tumor laterality in colorectal cancer, with right-sided tumors being associated with a worse prognosis compared to left-sided tumors [27]. In our cohort, right-sided tumors were significantly associated with poorer PFS (p = 0.05) and OS (p < 0.001) (Supplementary Figure S2C).

B7-H3 was observed as membranous staining on colorectal cancer cells in TMA sections from mCRC patients (Supplementary Figure 3A; Figure 1A). All replicates of tumor tissues from the n = 55 mCRC patients analyzed showed B7-H3 expression with a staining score of ≥1 (Figure 1B). A strong averaged staining intensity (staining score = 3) was observed in 43.6%, an intermediate averaged staining intensity (staining score = 2) was observed in 18.2%, and 10.9% of the cases were stained with a weak intensity (staining score = 1). Additionally, 14.6% of patients had a mean staining score of 2.5, and 12.7% had a score of 1.5 (Table 2). Notably, none of the analyzed mCRC patient tissues was negative for B7-H3. The mean percentage of positively stained cells was 95.0%, ranging from a minimum of 45.0% to a maximum of 100% B7-H3-positive cells. H-scores were predominantly elevated, with 100% of the metastatic colorectal cancer cases analyzed having an H-score ≥45. The median H-score was 230, with a range from 45 to 300 and an interquartile range (IQR) of 147.5 (Figure 1C). To further assess the consistency and pattern of B7-H3 expression beyond TMA sampling, whole slide imaging (WSI) was performed on tissue sections from three mCRC patient tumors. These analyses confirmed uniform B7-H3 expression across both primary tumors and metastatic lesions (Supplementary Figure S3B).

To analyze difference in clinical and pathological patient parameters depending on B7-H3 expression, patients were subgrouped in three tertiles by H-Score (1 = 0–190, 2 = 190–285, 3 = 285–300). No significant difference for age at diagnosis and gender was observed (p = 0.36 and p = 0.84, respectively). Next, parameters for the disease status at diagnosis were analyzed for differences regarding B7-H3 expression levels. No significant differences for these parameters were observed: primary tumor localization (p = 0.83), tumor grading (p = 0.23), T-stage (p = 0.80), N-stage (p = 0.87), metastases at diagnosis (p = 0.89), and lymphovascular invasion (L-stage) (p = 0.31). Furthermore, we investigated if the treatment has an impact on B7-H3 expression. The number of chemotherapy lines (p = 0.60) did not show any significant effect on tumor-B7-H3 expression (Figure 2A). In addition to patient characteristics, the clinical parameters of tumor markers (soluble, tumor-expressed and mutations) were associated with H-Score tertiles. For serum CEA level (p = 0.41), CA19-9 level (p = 0.81), HER2 score (p = 0.05), microsatellite status (p = 0.22), KRAS mutation (p = 0.09), NRAS mutation (p = 0.26), BRAF mutation (p = 0.49), PIK3CA mutation (p = 1.00), and p53 mutation (p = 0.22), no significant differences were observed between the three H-Score tertiles (Figure 2B).

In summary, no significant association was found between B7-H3 expression, stratified by H-score tertiles, and the clinical or pathological characteristics of patients with mCRC.

Next, we investigated if expression levels of B7-H3 have a potential impact on the outcomes of mCRC patients. To this end, we analyzed the association of B7-H3 H-scores with PFS and OS. A logrank test showed no significant difference in PFS across the three H-Score groups (p = 0.41), indicating that B7-H3 expression was not associated with PFS in this cohort (Figure 3A). Similarly, no significant differences in OS were observed between the H-Score tertiles (p = 0.97) (Figure 3B).

Finally, we investigated whether alterations of B7-H3 expression levels in tumor tissues of mCRC patients occurred after the progression of the disease. Therefore, longitudinal, sequential data from an additional biopsy were immunohistochemically analyzed from n = 7 patients. The mean time between the two biopsies was 19.71 months (SD = 6.55), with a range from 7 to 27 months. The analysis focused exclusively on metastatic lesions, considering sequential, subsequent specimens from the liver (n = 6) and the lung (n = 1) (Figure 4A). The range of B7-H3 H-scores was found to vary from 90 to 300 at the initial analysis, and from 200 to 290 for the sequential biopsies time (Figure 4B). It should be noted that the initial values represent the mean scores from two replicates, whereas the sequential biopsies were based on a single sample per patient. No significant difference in H-Scores was found between the two time points (paired Wilcoxon test, p = 0.81). H-Scores at both time points showed a moderate positive correlation (Spearman’s ρ = 0.56). Although not statistically significant (p = 0.19), the moderate correlation suggests a potential trend toward consistent B7-H3 expression over time (Figure 4C).

These results demonstrate that B7-H3 is expressed in mCRC on both primary tumor tissue and metastases and is not altered during progression of the disease.