Optimised ARID1A immunohistochemistry is an accurate predictor of ARID1A mutational status in gynaecological cancers

Abstract

ARID1A is a tumour suppressor gene that is frequently mutated in clear cell and endometrioid carcinomas of the ovary and endometrium and is an important clinical biomarker for novel treatment approaches for patients with ARID1A defects. However, the accuracy of ARID1A immunohistochemistry (IHC) as a surrogate for mutation status has not fully been established for patient stratification in clinical trials. Here we tested whether ARID1A IHC could reliably predict ARID1A mutations identified by next-generation sequencing. Three commercially available antibodies - EPR13501 (Abcam), D2A8U (Cell Signaling), and HPA005456 (Sigma) - were optimised for IHC using cell line models and human tissue, and screened across a cohort of 45 gynaecological tumours. IHC was scored independently by three pathologists using an immunoreactive score. ARID1A mutation status was assessed using two independent sequencing platforms and the concordance between ARID1A mutation and protein expression was evaluated using Receiver Operating Characteristic statistics. Overall, 21 ARID1A mutations were identified in 14/43 assessable tumours (33%), the majority of which were predicted to be deleterious. Mutations were identified in 6/17 (35%) ovarian clear cell carcinomas, 5/8 (63%) ovarian endometrioid carcinomas, 2/5 (40%) endometrial carcinomas, and 1/7 (14%) carcinosarcomas. ROC analysis identified greater than 95% concordance between mutation status and IHC using a modified immunoreactive score for all three antibodies allowing a definitive cut-point for ARID1A mutant status to be calculated. Comprehensive assessment of concordance of ARID1A IHC and mutation status identified EPR13501 as an optimal antibody, with 100% concordance between ARID1A mutation status and protein expression, across different gynaecological histological subtypes. It delivered the best inter-rater agreement between all pathologists, as well as a clear cost-benefit advantage. This could allow patients to be accurately stratified based on their ARID1A IHC status into early phase clinical trials.

Keywords: ARID1A immunohistochemistry; biomarker; clear cell ovarian cancer; next generation sequencing.

Figure 1

ARID1A antibody optimisation in tumour cell lines and tumour xenograft models. (A) Schematic of the ARID1A protein illustrating epitope regions of all three antibodies studied. The lollipop plot shows mutation loci of two cell lines used in antibody optimisation: HCT116 ARID1A isogenic mutant cell line Q456*/Q456*, nonsense mutation (depicted in orange); and TOV21G, a compound heterozygous cell line with mutation loci TOV21G p.548fs and p.756fs, frameshift mutations (depicted in mauve). (B) Immunoreactivity of ARID1A detected by all three antibodies in HCT116 ARID1A isogenic cell lines embedded in FFPE blocks. HCT116 –/– (ARID1A mutant) shows loss of ARID1A immunoreactivity whereas nuclear immunoreactivity was preserved in the HCT116 +/+ (ARID1A wild‐type) cell line. EPR13501, rabbit monoclonal, antigen retrieval using microwave and dilution 1:1000. D2A8U, rabbit monoclonal, antigen retrieval with pretreatment module and dilution 1:250. HPA005456, rabbit polyclonal, antigen retrieval with pretreatment module and dilution 1:400. Scale bar is equal to 100 µm. (C) ARID1A immunoreactivity in ovarian clear cell carcinoma cell lines. TOV21G (ARID1A mutant) shows loss of ARID1A immunoreactivity whereas immunoreactivity was preserved in the ES2 (ARID1A wild‐type) cell line. Scale bar is equal to 100 µm. (D) ARID1A immunoreactivity was detected with all three antibodies in xenograft models of HCT116 +/+ with loss of ARID1A expression in HCT116 –/– xenograft models. The background staining seen in cell lines was reduced in xenografts. Scale bar is equal to 100 µm.

Figure 2

Study workflow. Modified CONSORT diagram showing the processing of the 45 gynaecological cancer cases identified. After histopathology review, a representative block was chosen. The same FFPE block was used for extracting DNA for next‐generation sequencing (NGS) and IHC. One case failed NGS quality control and one case was not suitable for IHC analysis. Three histopathologists independently reviewed and scored cases. These were then averaged and compared to mutational status to establish concordance (n = 43).

Figure 3

Distribution of ARID1A mutations identified by targeted sequencing in gynaecological cancers. (A) A lollipop plot showing the distribution and classes of validated loss of function mutations in ARID1A detected by both sequencing platforms in 44 cases. Twenty‐one ARID1A mutations were identified in 14 patients. The majority were frameshift mutations (9, in mauve), with 11 nonsense mutations (orange) and 1 missense mutation (green). (B) Bar chart showing the frequency of ARID1A mutations according to histology. An enrichment of ARID1A mutations was seen in endometriosis‐related tumours and carcinosarcomas, with 63% of endometroid ovarian carcinomas (light blue), and 35% of OCCCs (dark blue) having an ARID1A mutation. Endometrial carcinoma includes ECC, EAE and DCE. (C) A VAF plot shows each case with the type of validated mutation (colour of circle) and coverage (size of circle) with corresponding protein expression on IHC (boxes below). Six cases had more than one mutation in ARID1A. Case 3705‐0541 was not suitable for IHC processing.

Figure 4

ARIDIA IHC shows good concordance with mutational analysis for all three antibodies in a variety of gynaecological cancers. (A) ARID1A immunoreactivity in grade 2 endometrioid adenocarcinoma of the ovary. Case 3705‐0460, with one mutation (p.Arg1505Ter), shows a lack of tumour cell staining with positive stromal staining (immunoreactive score with EPR13501: 1, D2A8U: 1 and HPA005456: 3). Case 3705‐0529, ARID1A wild‐type, shows positive tumour cell nuclear staining with an immunoreactive score with EPR13501: 12, D2A8U: 12 and HPA005456: 9. Scale bar is equal to 100 µm. (B) Histograms showing the distribution of all immunoreactive scores (n = 43) with all three antibodies, annotated with validated mutational status. The majority of ARID1A mutant cases (red) show low immunoreactivity scores and ARID1A wild‐type cases (blue) show high immunoreactivity scores. There is greatest variation in the HPA005456 scores.

Figure 5

ARIDIA IHC shows good concordance with mutational analysis for all three antibodies in OCCC. ARID1A immunoreactivity in OCCC. Case 3705‐0416, with two mutations (p.Gly191fs, p.Tyr485Ter), shows a lack of tumour cell staining with positive stromal staining (immunoreactive score with EPR13501: 0, D2A8U: 1 and HPA005456: 2). Case 3705‐0514, ARID1A wild‐type, shows positive tumour cell nuclear staining with an immunoreactive score with EPR13501: 12, D2A8U: 11, and HPA005456: 12. Scale bar is equal to 100 µm. (B) Histogram showing immunoreactivity scores for OCCC cases shows a bimodal distribution between mutant cases (loss of immunoreactivity) and ARID1A wild‐type cases (retention of immunoreactivity).