Prognostic Value of BRAF, Programmed Cell Death 1 (PD1), and PD Ligand 1 (PDL1) Protein Expression in Colon Adenocarcinoma

Patients with colorectal cancer in different stages show variable outcomes/therapeutic responses due to their distinct tumoral biomarkers and biological features. In this sense, this study aimed to explore the prognostic utility of BRAF, programmed death-1 (PD1), and its ligand (PDL1) protein signatures in colon adenocarcinoma. The selected protein markers were explored in 64 archived primary colon adenocarcinomas in relation to clinicopathological features. BRAF overexpression was found in 39% of the cases and was significantly associated with grade 3, N1, advanced Dukes stage, presence of relapse, and shorter overall survival (OS). PD1 expression in the infiltrating immune cells (IICs) exhibited significant association with T2/T3, N0/M0, early Dukes stage, and absence of relapse. PDL1 expression in IICs is significantly associated with advanced nodal stage/distant metastasis, advanced Dukes stage, and shorter OS. Meanwhile, PDL1 expression in neoplastic cells (NC) was associated with the advanced lymph node/Dukes stage. A positive combined expression pattern of PDL1 in NC/IICs was associated with poor prognostic indices. Tumor PDL1 expression can be an independent predictor of OS and DFS. The multivariate analyses revealed that short OS was independently associated with the RT side location of the tumor, PD1 expression in stromal IICs, and PDL1 expression in NC. In conclusion, overexpression of BRAF in colon adenocarcinoma is considered a poor prognostic pathological marker. In addition, PDL1 expression in NC is considered an independent prognostic factor for DFS/OS. Combined immunohistochemical assessment for BRAF and PD1/PDL1 protein expressions in colon adenocarcinoma might be beneficial for selecting patients for future targeted therapy.


Introduction
According to GLOBOCAN 2018 data, colorectal cancer (CRC) is the fourth most common cancer diagnosed globally [1,2]. Although cancer screening programs and improved demographic data, such as the patient's age, sex, tumor location, and post-operative course (recurrence and survival), were obtained from the patient's medical records.

Histopathology
Paraffin-embedded blocks of tumor tissue, as well as adjacent normal colon tissues from the patients, were retrieved from archives of pathology. Serial sections from each specimen were stained with H&E for histological evaluation. Two pathologists reviewed the histopathological features of each slide according to WHO classification [29]. TNM staging and Dukes staging of each tumor were reviewed according to Akkoca et al. [30].
The stain was interpreted independently by two observers blinded to the clinical outcome. The Hercept Test scoring system was used to detect the staining score of both PDL1 and PD1 in the infiltrating immune cells and PDL1 in tumor cells. For PDL1 expression of the tumor cells, the intensity of the stain was scored as 0 (no staining), 1 (light yellow), 2

Histopathology
Paraffin-embedded blocks of tumor tissue, as well as adjacent normal colon tissues from the patients, were retrieved from archives of pathology. Serial sections from each specimen were stained with H&E for histological evaluation. Two pathologists reviewed the histopathological features of each slide according to WHO classification [29]. TNM staging and Dukes staging of each tumor were reviewed according to Akkoca et al. [30].
The stain was interpreted independently by two observers blinded to the clinical outcome. The Hercept Test scoring system was used to detect the staining score of both PDL1 and PD1 in the infiltrating immune cells and PDL1 in tumor cells. For PDL1 expression of the tumor cells, the intensity of the stain was scored as 0 (no staining), 1 (light yellow), 2 (brown), and 3 (deep brown). The number of stained cells per 100 was scored as 1 (≤10%), 2 (10%~50%), and 3 (≥50%). High PDL1 expression was detected when the product of the staining strength score multiplied and the number of stained cells per 100 cells was no less than three. With regard to immune cell-specific PDL1/PD1 expressions, the percentage of stained cells per 100 cells were detected and categorized as 0-9%, 10-49%, and 50-100% stained immune cells [31].
For BRAF staining, the intensity of the anti-BRAF antibody in tumor cells was recorded on a 0-3 scale. The expression was mainly cytoplasmic with nuclear staining in cases with strong and moderate cytoplasmic staining. Strong cytoplasmic with or without nuclear staining was scored as 3, moderate cytoplasmic staining with or without nuclear staining as 2, weak cytoplasmic staining as 1, and the absence of staining was scored as 0. In addition, any nuclear staining and the percentage of tumor cells stained positive with anti-BRAF antibodies were recorded. The cases were scored as dysregulated BRAF protein expression if > 80% of tumor cells expressed diffuse uniform unequivocal strong or moderate cytoplasmic staining with or without nuclear staining. However, they were scored negative for dysregulated BRAF expression if they showed no staining or weak, cytoplasmic, non-granular, uniform staining (stain intensity <80%). The cases with staining of isolated tumor cells in a tumor and those who showed no staining were also considered negative cases for the dysregulated BRAF expression. The cases were scored as equivocal if they displayed ambiguous, heterogeneous, non-uniform cytoplasmic staining in tumor cells with or without nuclear staining [24].

Combined Expression Patterns of PDL1 in Neoplastic Cells (NC) and Infiltrating Immune Cells (IIC)
The study cases were categorized into four groups according to the combined expression patterns of PDL1 in neoplastic cells (NC) and infiltrating immune cells (IIC), a method which is validated by Valentini et al. 2018 (22). Group A (NC−/IIC−) which was negative in NCs and IICs; Group B (NC+/IIC−) which was positive only in neoplastic cells; Group C (NC−/IIC+) was positive only in IICs; Group D (NC+/IIC+) was positive both in NCs and IICs. The expression pattern of each marker in both NC and IIC was tested for association with clinicopathological parameters.

Statistical Analysis
Data were analyzed using R version 3.5.1 and SPSS version 23.0. Chi-square and Fisher's Exact tests were applied for qualitative variables (when appropriate), while student's-t and Mann-Whitney U tests were employed for continuous attributes. p-value ≤ 0.05 was set to be significant.

BRAF Protein Expression and Association with Clinicopathological Prognostic Factors of Colon Adenocarcinoma
BRAF protein expression was considered positive in 25 cases (39.1%). The staining was cytoplasmic with or without nuclear staining. Adjacent non-neoplastic colorectal tissue showed scattered nuclear staining in the mucosa and/or cytoplasmic staining in smooth muscle ( Figure 2).

BRAF Protein Expression and Association with Clinicopathological Prognostic Factors of Colon Adenocarcinoma
BRAF protein expression was considered positive in 25 cases (39.1%). The staining was cytoplasmic with or without nuclear staining. Adjacent non-neoplastic colorectal tissue showed scattered nuclear staining in the mucosa and/or cytoplasmic staining in smooth muscle ( Figure 2).
As regards the association of BRAF protein expression with the clinicopathological characteristics of the studied colonic adenocarcinoma cases, using a two-sided Chi-square test, there were significant associations with grade III (p = 0.04), N1 (p = 0.00), advanced Dukes stage (C-D) (p = 0.02), presence of relapse (p = 0.02), and shorter overall survival (OS; p = 0.00). At the same time, BRAF protein expression was not associated with the presence of lymphovascular invasion (LVI; p = 0.06) ( Table 2).  As regards the association of BRAF protein expression with the clinicopathological characteristics of the studied colonic adenocarcinoma cases, using a two-sided Chi-square test, there were significant associations with grade III (p = 0.04), N1 (p = 0.00), advanced Dukes stage (C-D) (p = 0.02), presence of relapse (p = 0.02), and shorter overall survival (OS; p = 0.00). At the same time, BRAF protein expression was not associated with the presence of lymphovascular invasion (LVI; p = 0.06) ( Table 2).

Expression of PD1 and PDL1 and Association with Clinicopathological Prognostic Factors of Colon Adenocarcinomas
In adjacent non-neoplastic colonic mucosa epithelial cells, PD1 and PDL1 proteins showed no staining. The pattern of PD1 and PDL1 expression in neoplastic cells was either focal or diffuse, with a predominance of the focal pattern, particularly along with the tumor-stromal interface. While in infiltrating immune cells (IIC), a diffuse positive pattern was prevalent. Among the 64 patients, 41 (64%) had PD1 overexpression in IIC (Table 3, Figure 3).  Figure 4).

Correlation between BRAF, PD1, and the PDL1 Protein Expressions
By testing the way of linking the expression pattern of the three markers using the Spearman Bivariate correlation test, BRAF protein expression positivity did not correlate with either IICs PD1, NC PDL1, IIC PDL1, or the combined pattern of PDL1. Only NC PDL1 and IIC PDL1 show a considerable positive correlation to the combined pattern of PDL1 expression (r: 0.591 and 0.895, respectively (p = 0.000 for both)) ( Table 5).

Discussion
The recently developed immunotherapeutic strategies have yielded remarkable clinical results in controlling tumor growth in many tumors. It showed the highest response in melanoma, renal cell carcinoma, non-small cell lung carcinoma, and microsatellite instability-high CRC [17,32]. The year 2017 witnessed the first US Food and Drug Administration (FDA) approval of immune checkpoint inhibitor (ICI) immunotherapy for the management of CRC [33]. However, PD1/PDL1 blockade therapy is significantly helpful only in a group of patients, and the others either show resistance or only respond transiently to this therapy [34]. So, identifying resistance mechanisms is crucial to enhance the reach of more responders to this therapy.
Also, accumulating evidence suggested that dysregulated BRAF expression has an immunosuppressive effect and a role in poor response to PD1/PDL1 checkpoint inhibitors targeted immunotherapy. In melanoma, there are mounting data that oncogenic BRAF contributes to immune escape, and several clinical trials combined BRAF inhibitors with immune checkpoint blockade [35]. For these considerations, IHC expression of PD1, PDL1, and BRAF proteins evaluation before therapy may help to determine patients that will benefit from immunotherapy and could be used as a base to design novel combination therapy (immunotherapy and BRAF inhibitors) for CRC.
BRAF is a protein kinase and part of the mitogen-activated protein (MAP) kinase signaling cascade, which involves the transduction of a growth signal from the cell membrane to the nucleus via a chain of protein kinases, and is responsible for cellular proliferation and survival [24]. Detection of BRAF in colon carcinoma has the potential as a prognostic marker and also as treatment target for new BRAF inhibitors, such as vemurafenib [35,36].
Genetic testing is expensive, with high-level laboratory requirements, and needs strict quality control and professional knowledge of molecular detection technology of PCR; hence, it is not conducive to clinical application in areas of limited resources. In contrast, IHC is economical, simple, and feasible. Some studies assessed the feasibility of IHC instead of PCR to detect the mutated BRAF and reported near-to-complete concordance between both techniques in various cancers, including colon carcinomas [37][38][39]. These results support using IHC as a simplified strategy to screen colorectal cancers in clinical practice [40].
Based on previous experience, we used immunohistochemistry to detect dysregulated protein BRAF expression. We followed the previous recommendation to consider diffuse cytoplasmic staining with moderate to intense staining in >80% of tumor cells as positive dysregulated expression; however, weak staining is negative, and heterogenous staining is equivocal [24]; however, using specific anti-BRAF monoclonal antibodies is recommended in future studies to uncover the mutational status of the tested samples.
BRAF protein expression was recorded in 25 cases (39.1%) in the current study with significant association with the advanced grade, lymph node state, Dukes staging, and the occurrence of relapse and short OS. From these results, we concluded that BRAF protein expression in CRC is a poor prognostic marker. These findings agree with the previous studies' results in which the authors explored the BRAF gene and protein expressions, and found that these expressions were associated with poor prognostic parameters, including the advanced grade and T stage of the tumor, and short OS in the case of protein expression [24,41]. Our findings support the previous studies, which report the significant implication of BRAF in cancer and associate right-sided colon cancer with worse clinical outcomes. BRAF protein is a valuable biomarker for identifying patients who may benefit from a more individualized course of therapy [24,42]. We noticed occasional cytoplasmic and nuclear immunostaining for BRAF in normal mucosa cells (Figure 2). These findings should be interpreted with caution, as incomplete specificity of the anti-BRAF polyclonal antibodies used in the current work and the potential cross-reactivity with other epitopes could play part in this finding, and warrant further confirmatory studies that apply more specific monoclonal antibodies to differentiate mutant vs. wild BRAF protein.
Regarding PD1, positivity in IICs was observed in 64% of cases. PDL1 was expressed in tumor cells in 53.1% of cases, and IICs in 51.6% of cases. Significant correlation between PD1 positivity in IICs and early T stage, negative LN stage, early-stage Dukes, M0, and absence of relapse. These data reflected that immune cell PD1 expression is significantly associated with good prognostic pathological parameters, which agreed with Berntsson et al., who concluded that immune-specific PD1 is significantly associated with lower T and M stages, whatever the location of the tumor, as their study related the side to the prognostic impact of PD1 and PDL1 expressions [43].
As regards PDL1 protein expression in NC and IIC, positive PDL1 expressions in NC showed a significant association with aggressive clinicopathological parameters (advanced nodal stage and Dukes). These data agree with previous studies [4,21]. Juneja et al. confirmed that NC PDL1 could inhibit the antitumor immunity by inactivation of CD8+ TC sensitive to PD1 signaling, and lead to an increase in the aggressiveness of the tumor [21]. Shen et al. concluded a significant association of NC PDL1 expression with advanced cancer stage and lymphatic invasion based upon a meta-analysis of 3481 patients included in 10 studies [4]. However, Berntsson et al. confirmed that neoplastic cell expression of PDL1 was significantly associated with younger age and highly differentiated tumors, but this was in the right-side colon cancer only and not on the left side or in the total cohort study [43].
The IIC PDL1 expression was significantly associated with advanced nodal stage, metastasis, advanced Dukes, and short OS. These results were inconsistent with the Berntsson et al. findings, which confirmed that immune cell expression of PDL1 was significantly associated with lower T, N, and M stages [43]. Such type of difference in the results could be attributed to a large number of Berntsson et al. samples (557 cases) compared to ours (64 cases), the different monoclonal antibodies used, and the different cutoff points of positivity. In addition, the degree of staining intensity was considered in the current work, but not in the Berntsson et al. study [43].
Regarding the pattern of PDL1 expression and tumor immune microenvironment, our study identified four subsets (NC−/IIC−, NC+/IIC−, NC−/IIC+, and NC+/IIC+). We found that combined expression pattern group D was significantly associated with advanced LN and Dukes stage. This classification helps the oncologist select the patient candidate for immune therapy, as the first pattern (NC−/IIC−) will not benefit from the therapy, and the last pattern (group D) will be the ideal candidate for checkpoint inhibitors.
The Spearman Bivariate correlation test revealed a positive correlation between BRAF protein expression scores and positivity. In addition, the test revealed that NC PDL1 is only significantly correlated to the combined pattern of PDL1 expression. However, IIC PDL1 showed a strong positive correlation to the combined pattern of PDL1. BRAF protein expression score or positivity did not correlate with IIC PD1, NC PDL1, IIC PDL1, or the combined pattern of PDL1. These results were consistent with the Berntsson et al. study, both on the correlation between BRAF expression and both PD1 and PDL1 protein expressions [43]. From these results, we can suggest that BRAF protein assessment could be helpful for patients with CRC arranging for combined therapy (immunotherapy and BRAF inhibitors).
Significant heterogeneity in survival outcome characterizes colonic cancer patients with dysregulated BRAF expression due to the complex, and still not entirely fully elucidated, interactions between the clinical, genetic, and epigenetic landscape of BRAF expression [44]. Our study suggested the importance of testing patients for PD1 and PDL1 along with BRAF protein expression evaluation.
Multivariate regression analysis using all clinical and pathological variables failed to detect PD1 (IIC), PDL1 (IIC), and BRAF protein as independent predictors for survival and relapse. However, tumor PDL1 expression can independently predict OS and DFS. This finding was consistent with a recent Wang et al. meta-analysis, in which the meta-regression showed that "PD-L1 expression played a significant role on poor CRC OS (HR = 1.95, 95% CI (1.92, 3.98)) and DFS (HR = 2.14, 95% CI (0.73, 4.52))" and could independently predict a poor CRC prognosis [45].
Also, the multivariate analysis revealed that the right-side location of the CRC could be a potential predictor for the short OS. This result is congruent with the Baran et al. report, which emphasized that CRC is not a single entity, but its pathogenesis and treatment response could depend on the anatomical location (i.e., RT vs. LT side). Patients with left-sided CRC showed more response to "5-fluorouracil (5-FU)-based regimes" as one of the adjuvant chemotherapies and also to "anti-epidermal growth factor receptor therapy" as targeted therapy with a better prognosis. In comparison, patients with right-sided CRC showed poor response to conventional chemotherapies, but demonstrated more promising results with immunotherapies, as it is characterized by an increase in the antigenic load [46].
It is noteworthy to consider the limited sample size in this study and the absence of applying a BRAF mutation-specific assay. In this sense, large-scale studies using highly specific (monoclonal antibodies) assays for IHC analysis for the detection of the BRAF mutation are highly recommended. In addition, the adoption of the recent "WHO classification of the digestive system tumors" is warranted in future studies.

Conclusions
The current study concluded that overexpression of BRAF protein in colorectal carcinoma is a poor prognostic pathological marker. In addition, PDL1 expression in NC is considered an independent prognostic factor for DFS and OS. Our study can suggest that combined immunohistochemical assessment for BRAF protein, PD1, and PDL1 expression in CRC could be beneficial for selecting patients for future combined immunotherapy and BRAF inhibitors. Informed Consent Statement: Not applicable, as the included specimens retrospectively were archived in the pathology lab.