Seventy-six colorectal cancer tumour blocks were identified and selected for the study. These cases were examined and reported by the Department of Pathology, St. John's Medical College and Hospital, Bangalore, India between 2013 and 2017. The present retrospective study was approved by the Institutional Ethical Committee, St John's Medical College and Hospital. Representative formalin-fixed and paraffin-embedded tumour blocks from each of the selected cases were obtained for the study. All tumours with a >50% tumour content, as estimated by a pathologist from the consecutive series, were chosen for analysis. Clinico-pathological characteristics, such as age, sex, grade, pathological and lymph node stage, histological type, lymphocytic response, lymphovascular invasion and tumour site were obtained from the clinical and histopathological records from the hospital.

TMA was constructed using the Quick Ray manual tissue microarrayer (Unitma Co., Ltd.). A master block grid plan was made with adequate precautions to ensure an accurate orientation and unambiguous specimen identification. Block construction was performed according to the manufacturer's instructions. Two cores of 1.5 mm each were taken from each tumour block. Sections were cut and stained with hematoxylin and eosin (H&E) to confirm the adequate representation of each tumour. Cores with <100 interpretable tumour cells were excluded from the analysis.

IHC was performed for CDX2, BRCA1 and MMR proteins (MLH1, PMS2, MSH2 and MSH6), according to standard procedures. Briefly, 5-µm thick sections were placed on poly-L-lysine-coated slides and subjected to deparaffinization in xylene and rehydrated in graded alcohol. Following blocking endogenous peroxidase with a 3% hydrogen peroxide solution, antigen retrieval was performed in 0.01 mol/l EDTA buffer at pH 8 using a heat triggered multi-epitope retrieval system (PathnSitu Biotechnologies Pvt. Ltd.) for 15 min at 90˚C. Primary blocking was done with 1% BSA (Sigma-Aldrich; Merck KGaA) for 30 min at room temperature. Details of the primary antibody clone, source and the dilutions are shown in Table SI. Primary antibodies were applied for 1 h at room temperature. Sections were further incubated with secondary antibody (cat. no. K5007; EnVision Detection System; Dako; Agilent Technologies, Inc.) for 20 min as per the kit instructions, followed by colour development using 3, 3-diaminobenzidine for 10 min. Sections were counterstained with haematoxylin for 5 min at room temperature and mounted after dehydration in graded alcohol and xylene. Appropriate positive and negative controls were run for each batch.

For CDX2 and BRCA1 scoring, each tumour core was scored for intensity as follows: 0=no staining; 1=weak staining; 2=moderate staining; 3=strong staining. The percentage of cells stained was estimated from 0-100%. The histochemical score (H score range, 0-300) was calculated by multiplying the intensity and the percentage of staining. Although BRCA1 staining was observed in both the nucleus and the cytoplasm, only its nuclear presence was evaluated, indicating BRCA1 functional ability. A nuclear H score of ≥10 was considered as positive BRCA1 and CDX2 expression. The H score was used for the analysis of both proteins to obtain a quantitative estimate with a broad dynamic range from 0-300(22). In the absence of any standard diagnostic criteria for proteins such as BRCA1 and CDX2, the H score method was followed for quantitation.

MMR proteins were scored only on the percentage of stained tumour cells, irrespective of staining intensity. Standard guidelines were followed where any presence of MMR proteins in the nucleus is considered adequate and acceptable to record as intact expression (23). Intact normal staining of non-tumour cells was considered as an internal positive control. Cases with ≥10% tumour cells showing nuclear staining were considered positive (intact expression). Weak focal nuclear stain in <10% of tumour cells was considered focal expression and a complete absence of nuclear stain in the presence of the positive internal control (lymphocytes and stromal cells) was considered negative (loss of expression). The dual loss of either MSH2 and MSH6 or MLH1 and PMS2, or the isolated loss of PMS2 was considered as an MMR-deficient group (dMMR). Intact/focal protein presence of either MSH2/MSH6 or MLH1/PMS2 was considered as an MMR-proficient group (pMMR).

H&E-stained microscopic sections of tumors were simultaneously evaluated by two trained pathologists to confirm the presence of above proteins in the tumor cells in comparison to normal cells (24). Each tissue microarray core was examined microscopically by the same pathologists independently to count ~1,500 cells in three different areas on the tissue section. The mean expression levels of both the observers were taken as final scores. In case of disagreement, the final score was determined by consensus after re-examination.

The methods used for nucleic acid extraction and RT-qPCR have been described in detail in our previous publication (25). In brief, two 20-µm sections taken from each patient's tumour block were deparaffinized using heat, and then subjected to overnight digestion using proteinase K (cat. no. 19133; Qiagen GmbH). Total RNA was then extracted using TRI reagent according to manufacturer's instructions (cat. no. T9424; Sigma-Aldrich; Merck KGaA). RNA quantification was performed using the Qubit RNA BR Assay kit (cat. no. Q10210; Invitrogen; Thermo Fisher Scientific, Inc.) on a Qubit 2.0 Fluorometer (cat. no. Q32866; Invitrogen; Thermo Fisher Scientific, Inc). Samples with a yield of ≥500 ng RNA and adequate transcript preservation to show amplification by RT-qPCR were used for subsequent experiments.

miRNA present in total RNA extracted as described above was converted to cDNA using stem-loop primers specific for the chosen miR, according to published protocols. Detailed methodology for quantification of miR using qPCR is provided in our previous publication (26). Briefly, 50 ng of total RNA was used for cDNA conversion using the TaqMan microRNA Reverse Transcription kit (cat. no. 4366596; Applied Biosystems; Thermo Fisher Scientific, Inc.). The expression levels of a selected set of DDR miRs (miR183, miR96 and miR182) were determined, along with the control miR (RNU48). The assay IDs (cat. no. 4427975; Applied Biosystems; Thermo Fisher Scientific, Inc,) and sequences of the miRs are shown in Tables SII and SIII. Cycle threshold (Ct) values for the test miRs were normalized relative to the mean Ct value of the control miR for each sample as ΔCt. The relative normalized units of expression of the test miRs were calculated as 15-ΔCt, representing the dynamic range of the assay as being 15 Cts.

Descriptive analysis was used to tabulate the clinical variables amongst the various groups segregated based on the presence or absence of the chosen protein markers. All markers were expressed as the proportion of cases noted as positive or negative. Parametric or non-parametric tests of significance, such as paired Student's t-test or the Mann Whitney U test, based on the normality of distribution, were applied to determine the levels of significance in the distribution of chosen variables between the groups. The correlation between protein expression levels was calculated using Pearson's correlation coefficient. In the absence of prior available data for combined loss of BRCA1 and MMR proteins, sample size estimates were not attempted. Analysis was performed on XLStat software (version 2019.4.2; Addinsoft). P<0.05 was considered to indicate a statistically significant difference.

The mean and median age of the CRC patients was 54.9 and 55 years, respectively. Of the 76 tumours studied, two were uninterpretable in CDX2 and BRCA1 IHC, one tumour yielded insufficient RNA and one tumour had loss of all MMR proteins, and therefore, only 72 (95%) tumours could be satisfactorily interpreted. A slight male preponderance was noted with 58% (42/72) in the group of tumours selected. The right and left sided tumours were equal in ratio. Most tumours belonged to stage II (38%) and stage III (43%). Details on the tumour grade were available for 70% (51/72) of tumours, and most (84%) of them were grade II tumours. Approximately 40% (29/72) of the tumours had a high lymphocytic response and lympho-vascular invasion was present in 43% (31/72) of tumours.

Of the 36 right sided tumors, 12 (33%) were mucinous tumors. The presence of the CDX2, BRCA1 proteins and the MMR status were further examined among mucinous and non-mucinous tumours. No statistically significant difference was observed between the two groups (data not shown).

CDX2 is known to be widely expressed in the large intestine (11) and showed nuclear staining with bright intensity. BRCA1 staining intensity was lower compared with CDX2 expression and varied across tumours. The intensity of MMR protein staining also varied widely.

Among the four MMR proteins studied, MSH2 and PMS2 had intact/focal expression in 89% of the tumours. Table I details the expression levels of all the MMR proteins. MLH1 protein was detected in a small proportion of tumours (42%). MSH6 was always intact (60%) in the presence of MSH2, while the presence of PMS2 was observed in 34/42 tumours when MLH1 was absent. Tumours with the dual loss of MSH2 and MSH6 (11%) had intact protein status for either MLH1 and/or PMS2, and vice versa (10%). The dual presence of MLH1 and PMS2 was observed in 29% of the tumours, while only one tumour had isolated loss of PMS2. A total of 22% (16/72) of the tumours were categorized as dMMR, as they had a dual loss of MSH2/MSH6 or MLH1/ PMS2, or the isolated loss of PMS2. Representative microscopic IHC images of all the MMR proteins showing positive nuclear stain are shown in Fig. 1A-D. As shown in Fig. 1, MLH1 presented with the lowest staining intensity compared to other MMR proteins. Table II shows the association of MMR status with clinical variables. None of the clinical variables were significantly different between the MMR proficient and deficient subtypes, aside from dMMR tumours, which had a higher lymph node spread (P-0.02). Among the dMMR tumours, 75% of them belonged to males and almost 70% of them belonged to stage III, although this was not statistically significant.

An intense nuclear staining of CDX2 protein in >10% of the tumour nuclei was considered positive and observed in 71% (51/72) of tumours (Fig. 1E; H score, 100x3-300). Clinical variables, such as age, sex, grade or stage did not differ between CDX2-positive and -negative tumours (Table II), although a higher proportion of right sided tumours was CDX2-negative (62% in right tumours vs. 45% in left tumours; P-0.2). Two-thirds (67%) of the CDX2-negative tumours belonged to males and more than half of the CDX-positive tumours were lymph node-negative.

BRCA1 nuclear expression was observed in only 21% (15/72) of the tumour samples (Fig. 1F; H score, 50x2-100). Cytoplasmic expression of BRCA1 was observed in an additional 14% (10/72) of tumours; however, they were considered negative. No significant differences were seen for any clinical variables between BRCA1-positive or -negative tumours, although a higher proportion of BRCA1-positive tumours (60%) was lymph-node negative (Table II).

Association of CDX2 loss is often reported to be high in MMR-deficient tumours (10,11). In the present study, 44% (7/16) of dMMR tumours showed loss of CDX2, compared with 25% (14/56) observed in pMMR tumours (P-0.1). However, BRCA1 and MMR were inversely correlated, with 38% (6/16) of the dMMR group expressing BRCA1 protein, compared with only 16% (9/56) in the pMMR subgroup (P-0.06). When the expressional pattern of BRCA1 and CDX2 were compared as H scores, a negative correlation was found between the two proteins (Pearson's correlation coefficient, -0.13; P-0.2), as expected.

Subsequently, the tumours were divided into two separate classes based on the dual presence and absence of BRCA1 and MMR status. While two-thirds of the tumours (53/72) showed either BRCA1 or MMR protein expression, only a small proportion of tumours (12.5%) had the combined presence of both. A small subset (10/72; 14%) of tumours was both BRCA1 negative and MMR deficient. When the clinical variables were compared between these groups (Table III), tumours with the combined loss of BRCA1 and MMR proteins showed aggressive features, such as a younger age, male preponderance and a higher proportion of grade III and stage III tumours. There was no difference in CDX2 expression between the two groups.

When the distribution of the miRs implicated in defective DDR (miR-183, miR-96 and miR-182) was examined in the two groups, tumours with the combined loss of BRCA1 and MMR showed trends of higher expression of all the three miRs (P-0.01, miR-182), as shown in Table III, indicating defective pathways for DNA damage repair in these tumours.