In colorectal cancer there is a need for molecular markers that can complement the histopathological staging in predicting the likelihood of disease recurrence following curatively intended surgery. B7-H3 is an immunoregulatory protein shown to be overexpressed in several cancer forms, often associated with more advanced disease and poor prognosis. We wanted to examine whether B7-H3 could be a potential prognostic marker in colorectal cancer. Paraffin-embedded samples from 277 colorectal cancer patients were immunostained with anti-B7-H3 antibody. B7-H3 was expressed in the tumor cell cytoplasm and cell membrane in 62% and 46% of the samples, respectively. Unexpectedly, B7-H3 was expressed in the nucleus in 30% of the tumors. The nuclear localization was confirmed by Western immunoblotting of subcellular fractions. Importantly, in colon cancer, nuclear B7-H3 expression was independently and significantly associated with reduced metastasis-free, disease-specific and overall survival. B7-H3 expression in tumor-associated vasculature and fibroblasts was observed in the majority of samples, and endothelial B7-H3 expression was also significantly associated with poor outcome in colon cancer. In rectal cancer patients, the only significant association was between fibroblast B7-H3 expression and shorter metastasis-free survival. Few significant associations to clinicopathological parameters were seen. The results indicate that nuclear B7-H3 might be involved in colon cancer progression and metastasis, and suggest that nuclear B7-H3 could become a useful prognostic marker in colon cancer.
Colorectal cancer (CRC) is globally the second most common cancer in women (570,000 cases/year) and the third most common in men (663,000 cases/year), accounting for 8% of all cancer deaths.1 CRC is classified according to the TNM system,2 but this purely anatomical staging system has its limitations. Thus, tumors with similar histopathological features can have significantly different clinical outcomes,3 probably reflecting biological and molecular heterogeneity not recognized by TNM staging. A significant proportion of patients with localized tumors, which in theory should be cured by surgical removal of the tumor, nevertheless develop lethal metastatic disease.4–6 These patients would most likely benefit from adjuvant chemotherapy, which is well established for TNM III patients but still remains controversial for node negative (TNM II) patients.4, 7, 8
To improve clinical outcome there is consequently a need for molecular markers that can complement the histopathological staging in predicting the likelihood of disease recurrence following curatively intended surgery. Such markers could thereby facilitate the risk-benefit assessment of adjuvant treatment, and also assist in identifying patients who need intensified postoperative surveillance. Several markers have been identified but none of them have so far been validated for clinical use.9
B7-H3, an immunoregulatory protein that belongs to the B7 family of T-cell co-regulatory molecules,10 may be a new potential prognostic marker. Its expression can be induced on T-cells, natural killer cells and antigen presenting cells.11, 12 Roth et al. found that it is also expressed in several nonhematopoietic organs, but not in colon epithelial cells.13 Functionally, B7-H3 was first described as a costimulatory molecule for T-cell activation and IFN-γ production.11 Several subsequent studies support a stimulatory immunological role of B7-H3,14, 15 also in the field of antitumor immunity.16, 17 In contrast to these findings, numerous other groups argue that B7-H3 has inhibitory immunological effects.12, 18, 19
It might be that the contrasting immunomodulatory effects of B7-H3 reflect the possible existence of two or more receptors with opposing functions. TLT-2 has been suggested as a potential receptor for B7-H3, binding to which enhances T-cell activation.14 However, this has subsequently been rejected by Leitner et al., who claim that B7-H3 is a potent inhibitor of human T-cell activation and that there is no evidence for B7-H3 and TLT-2 interaction.19
B7-H3 is overexpressed in several different cancer forms,20–26 including CRC,27 but the exact role of B7-H3 in cancer is still ambiguous. A few studies describe a positive correlation between tumor expression of B7-H3 and prognosis,20, 25 whereas most groups report that high tumor B7-H3 expression is associated with more advanced disease, increased risk of recurrence and/or shorter survival time.13, 21, 23, 24, 26, 28 Although a majority of studies on B7-H3 and cancer emphasize the immunological role of B7-H3, our group has shown that B7-H3 also affects chemosensitivity and cancer metastasis by interfering with signaling pathways activated in non-immunological systems.29–31
In the present study, we investigated the expression of B7-H3 in primary colorectal carcinomas in a large, unselected patient population and evaluated the association between B7-H3 expression and clinicopathological parameters and patient outcome.
Material and Methods
Colorectal cancer patients
Between September 1998 and July 2000, 316 patients from five hospitals in the Oslo region were included in the study at the time of primary surgery for assumed or verified CRC. The study was approved by the Regional Ethics Committee (#S-98080) and informed consent was obtained from the patients. Further details have previously been described.32 In study cohort I, 31 patients were excluded from statistical analysis for the following reasons: not invasive cancer (25), histology other than adenocarcinoma (5) and unknown stage of disease (1). In addition, sections were not obtainable in eight cases. The total study population thus was 277 patients, with 89 (32%) of the tumors localized in the rectum and 188 (68%) in the colon. In study cohort II, in which outcome was assessed, the same 31 patients were again excluded and in addition 43 patients were excluded due to distant metastases at the time of surgery (34), inadequate surgical margins (7) and preoperative chemoradiotherapy (2). The outcome study population included 242 patients in TNM stages I-III who had undergone curative surgery, and paraffin sections were available from 237 of these individuals. Mean patient age was 73 years (range: 35–98 years). One hundred and sixty-three (67%) of the tumors were localized in the colon and 79 (33%) in the rectum.
Follow-up data were obtained from consecutive reports from physicians at the participating hospitals. Additional details regarding follow-up data are specified by Boye et al.33
Sections of formalin-fixed, paraffin-embedded tissue were immunostained using the Dako EnVision method (Dako, Glostrup, Denmark). The sections were deparaffinized, rehydrated and antigen retrieved in Tris/EDTA buffer at 100°C for 15 min, and left in the buffer in 10 min after boiling. Following rinse in distilled water and Dako wash buffer, they were treated with 0.03% hydrogen peroxide for 5 min to block endogenous peroxidase activity, before incubation with goat polyclonal affinity purified anti-human B7-H3 antibody (R&D Systems, Minneapolis, MN), diluted 1:200, for 30 min at room temperature. The sections were then rinsed in Dako wash buffer, and incubated with mouse anti-goat IgG-B (Santa Cruz Biotechnology, Santa Cruz, CA), diluted 1:100, for 30 min. Following rinse in Dako wash buffer the sections were incubated with Labelled Polymer-HRP anti-mouse (Dako), for 30 min at room temperature, and then rinsed two times in Dako wash buffer before 10 min incubation with diaminobenzene. After rinsing they were counterstained with hematoxylin, rinsed, dipped briefly in a water bath containing some drops of ammonia, before dehydration and mounting in Diatex.
The dilutions were made with Dako Cytomation Antibody Diluent. Negative controls included replacement of the primary antibody with normal polyclonal goat IgG of the same subclass and concentration, and incubation of sections with goat polyclonal anti-B7-H3 antibody pre-absorbed with 10 μg/ml recombinant human B7-H3 (R&D Systems). Positive controls (sections from colorectal tumor tissue known to express high amounts of B7-H3) were included in all series. The immunostained sections were scored microscopically by the study pathologist (J.M.N.), without the knowledge of patient outcome. The number of stained tumor cells was semi-quantitatively estimated, and graded as 0 (negative), 1 (less than 10% positive cells), 2 (10–49% positive cells) or 3 (more than 50% positive cells). Membrane, cytoplasmic and nuclear staining in tumor cells were recorded as individual variables. Staining of tumor-associated fibroblasts and endothelial cells was also recorded.
Preparation of nuclear and cytoplasmic protein fractions from frozen tumor tissue
Sub-fractioning of nuclear and cytoplasmic components was performed as previously described.34 Briefly, frozen tumor tissue was thawed, minced and homogenized, and following filtration nuclear and cytoplasmic components were isolated by centrifugation (37,000 rpm, 30 min). The supernatant represented the cytoplasmic protein fraction. Portions of the nuclear fraction were hematoxylin-stained and microscopically evaluated to confirm the isolation of highly purified nuclei. The nuclear pellet was resuspended in lysis buffer, sonicated, left on ice for 1 hr and then centrifuged to remove cellular debris. Both fractions were stored at −70°C.
Preparation of whole cell protein lysate from frozen tumor tissue
Whole cell protein lysate was extracted from frozen tumor tissue which was cut into 10 μm slices; 15–18 slices were homogenized with Lysing Matrix D beads (MP Biomedicals LLC, Illkirch, France) using a FastPrep FP120 Cell Disrupter (Qbiogene, Illkirch, France), in ice cold lysis buffer containing 20 mM Tris-HCl, pH 7.5, 137 mM NaCl, 100 mM NaF, 10% Glycerol, 1% NP-40, 1 mM PMSF and 0.02 mg/ml each of leupeptin and pepstatin. Lysates were then centrifuged and the supernatant was collected as total protein lysate and stored at −80°C.
Western immunoblot analysis
Protein quantitation was done by BCA Protein Assay (Thermo Scientific Pierce, Rockford, IL) according to the manufacturer's instructions. Five to seventeen microgram of protein per lane was separated with 4–12% NuPage Novex Bis-Tris gels (Invitrogen, Oslo, Norway) and the NuPage MOPS buffer system (Invitrogen) according to the manufacturer's instructions, and transferred onto Immobilon-P membranes (Millipore, Bedford, MA).
After blockage of non-specific binding sites with 5% non-fat dry milk in Phosphate Buffered Saline (Lonza, Verviers, Belgium) supplemented with 0.5% Tween (PBST), blots were incubated for 1 hr at room temperature with goat polyclonal affinity purified anti-human B7-H3 antibody (diluted 1:250; R&D systems), mouse monoclonal anti-α-tubulin antibody (1:5000; Calbiochem, Merck, Nottingham, UK), and mouse monoclonal anti-lamin B antibody (1:500; Pierce Thermo Scientific). All dilutions were made with PBST with 5% non-fat dry milk. After washing with PBST to remove residual primary antibodies, the membranes were incubated with horseradish peroxidase (HRP)-conjugated appropriate secondary antibodies (Dako) for 1 hr at room temperature. After washing with PBST, the bands were visualized using Super Signal West Dura Extended Duration Substrate (Thermo Scientific Pierce).
The membranes were then stripped for 40 min, to remove previous antibodies, and incubated with anti-LAMP2 antibody (1:500; Santa Cruz) for 1 hr at room temperature. After washing, the membranes were incubated with HRP-conjugated anti-mouse secondary antibody (1:5000; Dako) for 1 hr at room temperature, and the bands were visualized as described above.
To evaluate the associations between B7-H3 expression and clinicopathological parameters we divided the nuclear, cytoplasmic and membrane B7-H3 grades of the tumors into negative and positive staining. The endothelial and fibroblast B7-H3 grades were divided into weak (Grades 0 and 1) and strong staining (Grades 2 and 3). The associations were tested using two-tailed Fisher's exact test or linear-by-linear association chi-square test.
Univariate survival analysis was performed according to the Kaplan–Meier method, and survival was compared using the log rank test. Multivariate analysis was conducted using the Cox proportional hazards regression model with backward, stepwise elimination of variables. Survival was measured from date of surgery until death for overall and disease-specific survival, and from date of surgery until diagnosis of distant metastasis for metastasis-free survival. Data analysis was performed using SPSS version 16.0 (SPSS, Chigaco, IL). p Values <0.05 were considered as statistically significant.
Clinical and pathological features for the total study population are summarized in Table 1. Mean patient age at the time of surgery was 70 years (range: 21–98 years). The majority of the patients was in TNM stage II (40%), had pT3 tumors (65%) and no lymph node metastasis (61%). Eighty-three percent of the tumors were intermediately differentiated, and 64% had intermediate lymphocyte infiltration. Most tumors did not have vascular or perineural invasion (77% and 89%, respectively). Perinodal growth was seen in 61% of the 109 patients with regional lymph node metastasis. The clinical and pathological features for the outcome study population were reported in Boye et al.33
Table 1. Baseline clinicopathological data of the total study cohort
B7-H3 expression in primary colorectal carcinomas
The immunohistochemical expression levels of B7-H3 are shown in Table 2. Of the 275 tumors analyzed for B7-H3 expression, 171 samples (62%) displayed cytoplasmic staining (Fig. 1A), whereas membrane staining (Fig. 1B) was observed in 126 (46%) tumors. Interestingly, we observed nuclear staining in 82 of the 275 (30%) samples (Fig. 1C). In total, 72% (198 of 275) of the samples showed either nuclear, cytoplasm and/or membrane staining. In the majority of samples we also observed B7-H3 expression in tumor associated vasculature (252 of 277) and fibroblasts (245 of 277; Fig. 1D).
Table 2. Immunohistochemical expression levels of B7-H3 in primary colorectal carcinomas
Presence of nuclear B7-H3 in tumor samples assessed by Western immunoblotting
To confirm the presence of B7-H3 in tumor cell nuclei found by immunohistochemistry we performed Western immunoblotting. Tumor tissue from patients 19 and 54, with Grades 1 and 2 nuclear staining, respectively, were selected for analysis of nuclear and cytoplasmic protein fractions, as well as whole cell protein lysate. LAMP2, lamin B and α-tubulin were used as markers for membrane,35 nuclear36 and cytoplasmic37 fractions, respectively, to evaluate the purity of the fractions. Figure 1E shows that the isolation of the two fractions was successful and confirms that B7-H3 was present in the nuclear as well as in the cytoplasmic protein fraction from both tumors. Furthermore, it displays that nuclear and cytoplasmic B7-H3 has similar molecular weight.
B7-H3 expression and association with clinicopathological parameters
Few significant associations were found between the immunohistochemical expression of B7-H3 in CRC (Table 2) and the clinicopathological parameters shown in Table 1. However, there was a weak positive association between strong endothelial expression of B7-H3 and perineural invasion, both in the whole population (p = 0.046; Supporting Information Table 1), and in colon cancer separately (p = 0.04, data not shown). We did not observe any differences in B7-H3 expression in colon versus rectal cancer, except for the higher frequency of strong B7-H3 fibroblast expression in rectal cancer samples (55 of 89) than in colon cancer samples (91 of 188; p = 0.04; Supporting Information Table 1). We found a significant positive association between nuclear B7-H3 expression and vascular invasion (p = 0.04; Supporting Information Table 2) in colon but not in rectal cancer patients, whereas for rectal cancer there was an association between both cytoplasmic and membrane B7-H3 expression and TNM-stage (both p = 0.02, data not shown). B7-H3 expression was not associated with tumor growth pattern, nor was there any difference in expression level between the infiltrative border and center of the tumor.
When we looked at the associations between the expression of B7-H3 in the different tumor cell compartments and in stromal cells, there was a striking difference between colon and rectal cancer patients regarding nuclear B7-H3. In colon cancer, nuclear expression of B7-H3 was strongly associated with cytoplasmic, endothelial and fibroblast B7-H3 expression (p = 0.002, 0.02 and 0.002, respectively, data not shown), but not with membrane B7-H3 expression. In rectal cancer, we saw the opposite; nuclear expression of B7-H3 was significantly associated only with B7-H3 membrane expression (p = 0.03, data not shown).
Patient outcome: Associations with clinicopathological parameters and B7-H3 expression
These analyses were performed on the outcome study population. The prognostic significance of clinical and pathological variables is shown in Boye et al.33
There were no significant associations between the expression of B7-H3 and outcome in CRC patients, except for the weak association between nuclear B7-H3 and reduced overall survival (p = 0.049, Table 3 and Fig. 2). Notably, when analyzing colon cancer patients separately, there was a strong association between nuclear B7-H3 and inferior patient outcome (Fig. 2 and Supporting Information Table 3). The 5-year metastasis-free survival rate was 54% for colon cancer patients with nuclear B7-H3 positive tumors, compared with 86% for patients with nuclear B7-H3 negative tumors (p < 0.001, log rank test), and the 5-year overall survival rate was 62% and 79%, respectively (p = 0.004, log rank test). We also found significant associations between endothelial B7-H3 expression and reduced metastasis-free, disease-specific and overall survival (p = 0.02, 0.02 and 0.03, respectively, Supporting Information Table 3). In rectal cancer patients, the only significant association was between fibroblast B7-H3 expression and shorter metastasis-free survival (p = 0.02, data not shown), whereas for nuclear B7-H3 there was a trend towards increased metastasis-free survival, although not statistically significant (p = 0.067, data not shown).
Table 3. Survival analyses of B7-H3 in colorectal cancer
To determine if the relationship between nuclear B7-H3 expression and poor outcome in colon cancer patients was independent of other clinical and pathological parameters, multivariate Cox regression analysis was performed. Variables included in the multivariate analysis were nuclear B7-H3, age, gender, TNM stage, differentiation, lymphocyte infiltration, vascular invasion and perineural invasion. Interestingly, nuclear B7-H3 was independently and significantly associated with metastasis-free survival (Table 4). It was also an independent prognostic factor for disease-specific survival [p = 0.005; hazard ratio 2.9; 95% confidence interval (CI) 1.4–6.3; data not shown] and overall survival (p = 0.007, hazard ratio 1.9; 95% CI 1.2–3.1; data not shown). Besides nuclear B7-H3, TNM stage was the only variable independently and significantly associated to metastasis-free (Table 4) and disease-specific survival (data not shown) in colon cancer patients. For overall survival, TNM and age was additional independent and significant prognostic factors (data not shown).
Table 4. Multivariate Cox regression analysis, colon cancer patients1
Although B7-H3 is structurally a type I transmembrane protein, most immunohistochemical studies on B7-H3 in cancer report that it is found in the tumor cell cytoplasm as well as in the membrane.13, 20, 24, 26–28, 38 In this prospective study, we demonstrate for the first time nuclear expression of B7-H3, with positive staining of tumor cell nuclei in 30% of colorectal tumor samples. The antibody used is a commercially available antibody known to specifically recognize B7-H3, and it has been used in several immunohistochemical studies.13, 21–24, 26, 28 We have further validated its specificity in negative controls where the primary antibody was (a) replaced with normal polyclonal goat IgG of the same subclass and concentration and (b) pre-absorbed with 10 μg/ml recombinant human B7-H3. The nuclear localization of B7-H3 found by immunohistochemistry was confirmed by Western immunoblotting, which also ascertained that nuclear and cytoplasmic B7-H3 had similar molecular weight. This is compatible with the presence of full-length B7-H3 protein in the tumor cell nuclei.
Importantly, nuclear expression of B7-H3 was strongly and independently associated with reduced metastasis-free, disease-specific and overall survival in colon cancer patients. The validity of our findings is strengthened by the prospective study design, ensuring an unbiased selection of patients, and the stage distribution and patient survival was similar to the results from a large population-based study consisting of more than 100,000 patients.4 This suggests that nuclear B7-H3 could become a useful prognostic marker in colon cancer.
It should be noticed that the outcome study population had relatively few rectal cancer patients (n = 79), and this might partly explain the observed contrasting prognostic value of nuclear B7-H3 in colon versus rectal cancer. However, there is accumulating evidence that colon and rectal tumors are distinct biological entities,39, 40 and similar findings have been reported for other biomarkers.39, 41, 42 Contradictory prognostic effects of B7-H3 in different cancer forms are known from previous studies. High tumor or tumor vasculature B7-H3 expression was associated with increased survival in gastric carcinoma20 and pancreatic cancer,25 whereas it was associated with reduced survival in prostate cancer,13, 21 clear cell renal cell carcinoma23 and ovarian carcinoma.28 The assumption that B7-H3 plays different roles in colon and rectal cancer is in line with the finding that nuclear B7-H3 was associated only with membrane B7-H3 in rectal cancer, whereas in colon cancer it was associated with cytoplasmic, endothelial and fibroblast B7-H3 but not with membrane B7-H3. Further support is provided by the significantly higher frequency of strong B7-H3 fibroblast expression in rectal cancer, and the prognostic significance of endothelial B7-H3 in colon but not rectal cancer. Altogether, these findings indicate that the interplay between the tumor cells and the microenvironment differs in colon versus rectal cancer, and the observed contrasting prognostic value of nuclear B7-H3 might at least partially be ascribed to this. Correspondingly, Nielsen et al. found that microRNA-21, although equally expressed in the stroma of colorectal cancers, was associated with poor prognosis only in colon cancer patients and not in rectal cancer patients.42
B7-H3 has a molecular weight of about 90–100 kDa, depending on the glycosylation pattern, and thus its passage through the nuclear envelope necessitates an active transport mechanism. That B7-H3 does not contain a classical nuclear localization signal does not exclude nuclear import, as it recently has become apparent that several alternative pathways for nuclear transport exists.43 As B7-H3 is structurally a transmembrane protein it is likely that endocytosis is involved in its nuclear translocalization, in line with findings on nuclear trafficking of proteins in the EGFR family and other membrane proteins.44
Although nuclear B7-H3 has not previously been described, Ghebeh et al. found that B7-H1, another transmembrane protein in the B7 family, was translocated from the membrane to the nucleus in cells treated with doxorubicin.45 Interestingly, their findings indicate that nuclear translocation might be essential for the anti-apoptotic effects of B7-H1. The importance of aberrant regulation of nuclear transport in transformation and oncogenesis has become increasingly evident in the past few years,46 and nuclear expression of other membrane proteins, including the 170 kDa epidermal growth factor receptor (EGFR), have been shown in various tumor tissues.47, 48 EGFR nuclear expression has also been associated with reduced survival.48, 49
The strong association found between nuclear B7-H3 and poor outcome in colon cancer, combined with the prognostic irrelevance of cytoplasmic and membrane B7-H3 implies that subcellular localization determines the functional activity of the protein. Similarly, distinct prognostic implications have been shown for cytoplasmic and nuclear ERβ2 immunoreactivity in breast cancer.50 Our findings suggest that nuclear B7-H3 might be involved in tumor progression, and this is supported by the significant association between nuclear B7-H3 and vascular invasion seen in colon cancer. Both our group29, 31 and other groups21, 24, 26 have previously proposed a link between B7-H3 and metastasis, although not related to subcellular localization.
In conclusion, we have demonstrated nuclear expression of B7-H3 in about one-third of colorectal carcinomas. Importantly, nuclear B7-H3 expression in colon cancer constitutes an independent predictor of disease recurrence following curatively intended surgery. This could indicate that nuclear B7-H3 is involved in colon cancer progression and metastasis, and the underlying mechanisms whereby nuclear B7-H3 exerts its activities will be explored in future experimental studies. If our findings can be reproduced in validation studies, nuclear B7-H3 status might be considered in colon cancer treatment decisions.
The authors thank Ellen Hellesylt for excellent technical assistance. The present study was supported by the Faculty of Medicine, University of Oslo, The Norwegian Cancer Society and The Norwegian Research Council.