Genomic alterations of the ARID1A, ARID1B and ARID2 genes were explored using The Cancer Genome Atlas-colon adenocarcinoma (TCGA-COAD) dataset through the Cancer Virtual Cohort Discovery Analysis Platform (CVCDAP) (https://omics.bjcancer.org/cvcdap/) (27). The expression correlations among the ARID1A, ARID1B and ARID2 genes in TCGA-COAD dataset were assessed using the Pearson correlation coefficient in the Gene Expression Profiling Interactive Analysis 2 platform (28). The promoter methylation levels of the ARID1A, ARID1B and ARID2 genes in TCGA-COAD dataset were examined through The University of ALabama at Birmingham CANcer data analysis portal (https://ualcan.path.uab.edu/) (29,30). The Kaplan-Meier (KM) plotter database (https://kmplot.com/analysis/) (31) was used to evaluate the association of ARID1A (Affymetrix probe ID: 218917_s_at; n=1,061), ARID1B (Affymetrix probe ID: 238043_at; n=814) and ARID2 (Affymetrix probe ID: 225486_at; n=814) with overall survival (OS) in patients with CRC. Low- and high-expression groups were determined using the ‘Auto select best cut-off’ option, which is based on the median expression values. The hazard ratio and log-rank P-value were automatically computed by the database for survival analysis. The KM plots for overall survival were generated directly by the KM plotter tool without any additional statistical analysis or modification. A late-stage crossover was observed in the survival curves, as provided by the tool.

The present study was approved by the Human Research Ethics Committee of Sawanpracharak Hospital (Nakhon Sawan, Thailand; certificate of approval no. 53/2567) and the Naresuan University Human Research Ethics Committee (Phitsanulok, Thailand; approval no. P1-0107/2567; certificate of approval no. 139/2024), and conducted in accordance with the principles of the Declaration of Helsinki. Tissue biopsies from 63 patients diagnosed with CRC of varying pathological differentiation were submitted to the Pathology Unit at Sawanpracharak Hospital between 2017 and 2021. This patient cohort included 27 males and 36 females, with a median age of 66 years (range, 52-97 years). Formalin-fixed, paraffin-embedded (FFPE) blocks containing the CRC tissues, including both cancerous and adjacent non-cancerous areas, were obtained for each patient. The clinicopathological data, including age, sex, tumor location, tumor size, pathological differentiation, American Joint Committee on Cancer (AJCC) staging 8th edition (32), tumor invasion, metastasis, lymphovascular invasion, comorbidities and follow-up period post-operation, were comprehensively assessed by a clinical pathologist. The exclusion criteria included patients with incomplete data, patients diagnosed with hereditary CRC syndromes, those with cancer of unknown primary origin and cases where a pathologist or researcher was unable to clarify the findings of the histological and/or immunohistochemical investigation. To ensure anonymity, each FFPE block was labeled with a unique research code and sensitive patient information was carefully protected.

To assess the expression of the ARID proteins, IHC was performed using the following specific antibodies: Anti-ARID1A rabbit polyclonal antibody (1:400; cat. no. HPA005456; MilliporeSigma), anti-ARID1B mouse monoclonal antibody (1:200; cat. no. ab57461; Abcam) and anti-ARID2 rabbit polyclonal antibody (1:250; cat. no. ab113283; Abcam). CRC tissue samples were initially fixed in 10% neutral buffered formalin (NBF), and then processed into FFPE blocks. These blocks were sectioned into 3-µm-thick slices, followed by deparaffinization in xylene and rehydration using an ethanol gradient. Antigen retrieval was achieved at 97˚C for 35 min using the heat-induced epitope retrieval method in citrate buffer (pH 6.0). Endogenous peroxidase activity was blocked with 3% hydrogen peroxide/sodium azide (NaN₃) for 25 min at room temperature (RT). After washing the slides three times with PBS (5 min each), non-specific protein binding was blocked with 0.1% NaN₃ for 20 min at RT. The slides were then incubated with the specified primary antibodies at 4˚C overnight in a humidified chamber. As a negative control, the primary antibody was replaced with PBS. Subsequent steps involved incubation with biotinylated goat anti-rabbit IgG (H+L) (from the Rabbit specific HRP/DAB Detection IHC Kit; cat. no. ab64261; Abcam) or goat anti-mouse IgG (H+L) secondary antibody [from the Mouse-specific HRP/DAB (ABC) Detection IHC Kit; cat. no. ab64259; Abcam] for 15 min at RT, followed by three washes with PBS (5 min each). The slides were then incubated with streptavidin peroxidase for 15 min at RT. After three additional washes with PBS, immunostaining was performed using the chromogen 3,3'-diaminobenzidine (DAB) substrate (1:50; cat. no. ab64238; Abcam) for 4 min at RT to detect specific antigen-antibody interactions, with the DAB reaction halted using distilled water. The slides were counterstained with Mayer's hematoxylin (C.V. Laboratories Co., Ltd.) by 3 dips at RT, washed in running tap water for 5 min, dehydrated with increasing concentrations of ethanol, cleared with xylene, mounted with mounting media (Permount; Thermo Fisher Scientific, Inc.) and covered with a cover slip.

In total, five independent areas per slide, covering both cancerous and adjacent non-cancerous regions in the same CRC tissues, were analyzed using the ZEN program (Rushmore Precision Co., Ltd.) and an Axiocam 105 color ZEISS microscope (Carl Zeiss AG) at high power fields with x40 magnification. ImageJ software (version 1.53c; National Institutes of Health) was employed to detect and analyze the ARID-positive cells. All images were evaluated in a double-blind manner by both a pathologist and the investigators. IHC scoring utilized the modified histoscore (H-score), which combines staining intensity with the percentage of positively stained cells to assess the abundance and distribution of proteins in tissue samples (33). Staining intensity was classified as follows: Negative staining (0), weak positivity (1), moderate positivity (2) and strong positivity (3) (34). The H-score was calculated using the following formula: H-score=[(0 x % negative cells) + (1 x % weak positive cells) + (2 x % moderate positive cells) + (3 x % strong positive cells)] (33). The H-score values ranged from 0 to 300. The ARID protein expression levels were categorized into two groups based on the median value: Low (< median value) and high (≥ median value). This classification approach aligns with the previously established optimal cut-off for SWI/SNF component expression (35).

Descriptive statistics are expressed as the mean ± SD and the median. Quantitative data are presented as the mean ± SEM. Comparisons between two groups were performed using the Mann-Whitney U test. The correlation among the protein expression levels of ARID1A, ARID1B and ARID2 were determined using Spearman's rank correlation coefficient. Pearson's χ² test was used to analyze the association between the ARID1A, ARID1B and ARID2 protein expression levels and the clinicopathological characteristics of patients with CRC when the expected value in <20% of cells was <5. In cases where this condition was violated in a 2x2 table, Fisher's exact probability test was performed (36). The relationship between ARID protein expression and progression-free survival (PFS) was assessed using KM analysis and the log-rank test. Moreover, the PFS univariate and multivariate analyses were performed using Cox proportional hazards regression analysis. All statistical analyses were performed using IBM SPSS statistical software (version 25; IBM Corp.) and GraphPad Prism 9 (Dotmatics). P<0.05 was considered to indicate a statistically significant difference.

Gene mutations in ARID1A, ARID1B and ARID2 in TCGA-COAD dataset were investigated using the CVCDAP platform. The analysis revealed that ~20.75% of patients with CRC had mutations in at least one of these ARID genes. Among these genes, ARID1A was the most prevalent, accounting for 13% of all samples. Mutations in ARID1B and ARID2 were each identified in 8% of the samples. Frameshift mutations were the most common in ARID1A, while missense mutations were the most frequently observed in both ARID1B and ARID2 (Fig. 1A). Additionally, significant positive correlations were found among the expression of ARID1A, ARID1B and ARID2. A strong correlation was found between ARID1A and ARID1B expression (r=0.71, P<0.001), while moderate correlations were observed between ARID1A and ARID2 (r=0.48, P<0.001) as well as between ARID1B and ARID2 (r=0.43, P<0.001). Pearson correlation analysis revealed linear relationships between gene expression pairs, as visualized in the scatter plots with clear positive trend lines in each panel (Fig. 1B-D). Furthermore, promoter methylation analysis revealed that all three ARID genes exhibited significantly higher methylation levels in the COAD samples compared with the normal tissues (Fig. 1E-G). These findings suggest that genetic mutations and epigenetic regulation may contribute to the altered expression of ARID1A, ARID1B and ARID2 in CRC.

A total of 63 patients diagnosed with CRC were included in the present study, with ages ranging from 52 to 97 years, yielding a mean age of 67.17±8.83 years and a median age of 66 years. Among the cohort, 36 patients (57.14%) were women and 27 (42.86%) were men. The tumor location was predominantly in the rectum/sigmoid colon (47.62%), followed by the right-sided colon (38.09%) and the left-sided colon (14.29%). Tumor sizes varied from 2.00 to 12.50 cm, with a mean size of 5.52±2.07 cm and a median size of 5.00 cm. Pathological differentiation examination indicated that the majority of tumors were well-differentiated adenocarcinomas, found in 42 patients (66.67%), followed by moderately differentiated tumors in 15 patients (23.81%) and poorly differentiated tumors in 6 patients (9.52%). According to the AJCC staging, patients were classified as Stage I (12.17%), Stage II (26.98%), Stage III (36.51%) and Stage IV (23.81%). The depth of tumor invasion examination indicated that 80.95% of cases were categorized as late-stage (pT3-pT4). Lymph node involvement (pN) was absent in 58.73% of patients (pN0), while 41.27% had one or more positive lymph node (pN1-pN2). Distant metastasis (pM1) was identified in 23.81% of the cohort. Additionally, nearly half of the patients (49.21%) exhibited lymphovascular invasion. Lymph node metastasis was detected in 26 of the 63 patients (41.27%). Furthermore, 73.02% of patients had comorbidities, including diabetes mellitus, hypertension and dyslipidemia, highlighting the complexity of the patient population. The clinicopathological characteristics of the 63 patients with CRC are summarized in Table I.

The expression levels of ARID1A, ARID1B and ARID2 in the cohort of 63 patients with CRC were analyzed using IHC. The results indicated that nuclear ARID1A, ARID1B and ARID2 proteins were predominantly found in the colonic epithelial cells that form the intestinal glands. Strong nuclear expression of these ARIDs was observed in the intestinal cells of adjacent non-cancerous areas, while cancerous regions exhibited weaker staining. Additionally, ARID2 also exhibited cytoplasmic localization (Fig. 2A-C).

Semi-quantitative analysis (Fig. 2D) revealed a significant decrease in ARID1A protein expression in cancerous areas (90.89±6.67) compared with adjacent non-cancerous areas (234.23±7.40) in all CRC cases (P<0.001). In well-differentiated tumors, ARID1A protein expression was significantly lower in cancerous areas (89.92±9.69) compared with adjacent non-cancerous areas (234.59±8.84) (P<0.001). Similarly, moderately differentiated cancerous areas exhibited reduced ARID1A levels (96.94±10.40) compared with adjacent non-cancerous areas (224.95±17.74) (P<0.001). Poorly differentiated tissues also displayed decreased ARID1A protein expression in cancerous areas (82.58±11.79) compared with adjacent non-cancerous areas (254.88±17.69) (P<0.01).

ARID1B protein expression was significantly decreased in cancerous areas (83.87±8.04) compared with adjacent non-cancerous areas (117.34±8.13) in all CRC cases (P<0.05). Well-differentiated cancerous areas had lower ARID1B levels (68.46±9.20) than adjacent non-cancerous areas (106.25±9.08) (P<0.01). However, no significant differences were observed in tissues with moderate or poor differentiation grades, although ARID1B expression tended to be lower in cancerous areas (P=0.351 and P=0.818, respectively) (Fig. 2E).

ARID2 expression was significantly lower in cancerous areas (50.51±4.43) than in adjacent non-cancerous tissues (114.26±14.45) in all CRC cases (P<0.001). In well-differentiated tumors, ARID2 expression was significantly lower in cancerous areas (51.83±5.80) compared with adjacent non-cancerous areas (104.44±6.11) (P<0.001). Moderately differentiated cancerous areas also had decreased ARID2 expression (43.94±8.94) compared with adjacent non-cancerous areas (85.66±12.53) (P<0.05). Similarly, ARID2 expression in poorly differentiated cancerous areas was significantly lower (57.69±6.02) than in adjacent non-cancerous areas (111.96±13.39) (P<0.01) (Fig. 2F).

Next, the H-score for the cancerous areas was used to categorize ARID expression as either low or high based on the median cut-off value. For ARID1A, scores <164 were categorized as ‘low expression’ and scores of ≥164 as ‘high expression.’ There were 57 cases with low ARID1A expression (90.48%) and 6 cases with high ARID1A expression (9.52%). The median H-score of ARID1B expression was 100, with 34 cases showing low ARID1B expression (53.97%) and 29 cases showing high ARID1B expression (46.03%). For ARID2, the median H-score was 78, with 15 cases showing high ARID2 expression (23.81%) and 48 cases showing low ARID2 expression (76.19%) (Fig. 2G).

The Spearman's correlation coefficient was used to assess the linear correlation among the expression levels of ARID1A, ARID1B and ARID2 in CRC tissues (Fig. 3). The results showed a moderate correlation between the H-scores of ARID1A and ARID1B (ρ=0.350, P=0.005), but no significant correlation between ARID1A and ARID2 (ρ=0.189, P=0.139) (Fig. 3A and B). Additionally, ARID1B expression was moderately correlated with ARID2 expression (ρ=0.436, P=0.0004; Fig. 3C). The ρ values of the correlation coefficients from the Spearman's correlation analysis are presented in a heatmap (Fig. 3D).

Fisher's exact and χ² analysis revealed that low ARID1A expression in patients with CRC was significantly associated with late-stage disease (P=0.049), a higher pN stage (P=0.038), the pM stage (P=0.025) and LNM (P=0.038) compared with those exhibiting high ARID1A expression. However, no significant differences were observed in age, sex, tumor location, tumor size, pathological differentiation, pT stage, lymphovascular invasion or comorbidities between the two expression groups. Furthermore, low ARID1B expression in patients with CRC was significantly associated with pathological differentiation (P=0.004), late-stage disease (P=0.038), pN stage (P=0.010) and LNM (P=0.010) compared with those exhibiting high ARID1B expression. However, ARID2 expression did not show a significant association with the clinicopathological characteristics (Table II). After applying the Bonferroni adjustment (adjusted P=0.0014) to all ARID associations, none of these associations remained statistically significant.

The associations between the 5-year PFS of patients and the expression levels of ARID1A, ARID1B and ARID2 were analyzed using KM curve and log-rank test analysis (Fig. 4A-C). The results revealed that patients with CRC exhibiting high ARID1A expression had a significantly shorter PFS time compared with those exhibiting low ARID1A expression (P=0.036). Additionally, there was a trend towards a shorter PFS time in patients with low ARID1B and ARID2 expression compared with those with high expression; however, the differences were not statistically significant.

Cox proportional hazards regression analysis was conducted to assess the significance of potential prognostic factors in patients with CRC. Univariate analysis revealed that low ARID1A expression (P=0.005) was significantly associated with PFS. Furthermore, the multivariate analysis, which included ARID1A, ARID1B and ARID2 expression as well as LNM status, indicated that low ARID1A expression (P=0.015) was an independent prognostic factor related to PFS (Table III).

To further assess the prognostic significance of ARID1A, ARID1B and ARID2, additional analysis was conducted using the KM plotter database. The results demonstrated that lower expression of all three ARID genes showed a general trend towards a shorter OS time. However, only ARID1A expression showed prognostic significance, with patients with CRC exhibiting low ARID1A expression having a significantly shorter OS time compared with those exhibiting high expression (Fig. 4D-F).