Application of ARID1A to murine formalin-fixed paraffin embedded tissue using immunohistochemistry

ARID1A is a known suppressor of tumour formation and the Human Protein Atlas antibody HPA005456 has been demonstrated in previous literature to stain tumour tissue by immunohistochemistry (IHC) in formalin-fixed paraffin embedded human tissue and human cell lines. This article details the validation of this antibody for immunohistochemistry of formalin-fixed paraffin embedded murine tissue using a Leica BondMax immunostainer. Using Western blot and IHC on murine wild-type and knockout tissue we have demonstrated that this antibody to ARID1A correctly stains murine tissue by immunohistochemistry.


Introduction
ARID1A (AT-rich interactive domain 1a) is a member of the SWI/ SNF family and its loss has been implicated as a factor in multiple premalignant and malignant conditions, including Barrett's oesophagus and oesophageal carcinoma as well as endometrial and clear cell ovarian carcinomas and their precursor endometriotic lesions [1][2][3][4] . The ARID1A antibody from Human Protein Atlas is a rabbit antibody generated against a PrEST (Protein Epitope Signature Tag) fragment of the ARID1A gene and affinity purified against the same fragment 5 . It is thus designated as being "monospecific" in that the affinity purification removes any non-specific or low affinity binders to the peptide. Through the Human Protein Atlas, the antibody has been tested on a wide variety of human tissue types and human malignancies, as well as for expression in immunofluorescence on U-2 OS, A-431 and U-251 MG cell lines. This demonstrates a nuclear expression in all cell lines and in the majority of tissue types 6 . However, the Western blot data were not supportive and did not produce staining corresponding to the expected size, although the data from a protein array did confirm a peak at the expected size. The antibody has been used to stain human colorectal cancers 7 , clear cell carcinomas 8 and on a variety of clear cell cancer cell lines 9 by immunohistochemistry.
To our knowledge, whilst the sequence homology between mouse and human ARID1A is 95%, this antibody has not been qualified using knockout tissue and has not been tested or published on murine tissue and this work represents the first data to do so.

Reagent details
Details of all reagents with reference to the immunohistochemical staining procedure can be found in Table 1.

Antibody details
Anti-ARID1A is a monospecific rabbit polyclonal generated to a PrEST sequence -PGLGNVAMGPRQHYPYGGPYDRVRTEPG IGPEGNMSTGAPQPNLMPSNPDSGMYSPSRYPPQQQQQQ QQRHDSYGNQFSTQGTPSGSPFPSQQTTMYQQQQQNYK (Table 2). The homology of the PrEST sequence used as immunogen

Amendments from Version 1
Following reviewers comments, the authors have modified the manuscript to reflect their views. The no primary antibody (NPA) data has been added to Figure 3 and an additional datafile added for the whole slide image. The authors feel that the persistant stromal staining demonstrates a limitation in the Tamoxifeninduced knockout model and not a non-specific binding of the antibody. This has been explained further in the conclusions section of the article. Donkey anti-rabbit biotin (Jackson Immunoresearch, Table 2) is specific for Rabbit IgG (Heavy and Light chains) and was affinity purified to remove cross-reactions to Bovine, Chicken, Goat, Guinea Pig, Syrian Hamster, Horse, Human, Mouse, Rat and Sheep. All slides were stained with a concentration of 4.8 µg/ml.

REVISED
Anti-GAPDH was used as a loading control for Western blots and was a rabbit monoclonal (Cell Signaling, Table 2). Detection antibody for the Western blot for ARID1a was Goat anti-rabbit IR Dye 680LT (Li-Cor Biosciences, Table 2) used at a concentration of 0.1 µg/ml and detection of GAPDH was Goat anti-rabbit IR Dye 800CW (Li-Cor Biosciences, Table 2).

Tissue details
All tissues and cell pellets used during the validation were fixed in Neutral Buffered Formaldehyde as specified (Table 3) before being transferred directly to 70% ethanol for no longer than 3 days. Tissue processing was conducted on a Leica ASP300 through graded ethanols before clearing in xylene and impregnation in molten paraffin wax (Fisher). All tissue sections were cut on a Leica rotary microtome at 3 µm.
Arid 1a -/mice were created by crossing Floxed Arid1a mice with ROSA26 Cre-ERT2 mice and resultant genotyping. Loss of Arid1a expression is expected following intraperitoneal injection of Tamoxifen.
Floxed Arid1a mice were a gift from Dr. Peri Tate, Sanger Institute, Hinxton UK; ROSA26 Cre-ERT2 mice were a gift from Prof Chambon, IGBMC, France. ES-2 cells were purchased from ATCC and RMG-II were a gift from Prof Huntsman, British Columbia Cancer Agency, Vancouver, Canada.

Experiment details
Western blot. Protein was extracted from the two clear cell carcinoma cell lines using a Tris-EDTA lysis buffer and run on a nondenaturing 3-8% Tris-acetate gel (Life Technologies). Following electrophoresis, the transfer membrane was probed with 0.2 µg/ml of anti-rabbit ARID1A (HPA005456) at 4°C overnight and 0.1 µg/ml anti-GAPDH (14C10) for the same length of time.
Detection of the anti-rabbit ARID1A was with Goat anti-rabbit IRDye 680LT (Li-Cor Biosciences) and the GAPDH was with Goat anti-rabbit IRDye 800CW (Li-Cor Biosciences) both at 0.1 µg/ml.
The antibody was validated on a Leica BondMax instrument using a Leica Intense R kit to a standardised in-house protocol. All reagents were from Leica as part of the Intense R kit and were conducted at room temperature, unless otherwise specified. All staining steps included individual washes in Leica Bond Wash after each step, as part of the protocol (Table 4). A full protocol for the validated conditions can be found in the supplementary material. In this protocol, the step named "primary" refers to the anti-ARID1a primary antibody.
A slide using the same conditions and retrieval but omitting the primary antibody was used to control for any background staining due to the retrieval and detection steps.

Results
To determine the correct cell line to utilise and to confirm the equivocal Western blot data from Human Protein Atlas, the antibody was used to stain a Western blot of two cell lines; ES-2 and RMG-II, both of which are cell lines derived from clear cell carcinoma and have been previously demonstrated as ARID1a wild-type and mutated, respectively 9 . It could be demonstrated that the HPA ARID1A antibody showed positive expression in ES-2 cell lines at the expected size of 270 kDa and no staining for RMG-II. The loading control of GAPDH showed that there were no loading issues ( Figure 1). Thus, these cell lines were chosen to be grown, formalin fixed and processed into paraffin wax for immunohistochemical validation.
For immunohistochemical validation, ES-2 and RMG-II cell lines were stained using three antigen retrieval conditions; ER1 (Sodium Citrate, pH6), ER2 (Tris/EDTA, pH9) and Enzyme 1 (Proteinase K, 100 µg/ml) at a fixed antibody concentration of 1 µg/ml. The enzyme retrieval demonstrated no nuclear signal for either ES2 or RMG-II cell pellets and was discarded for future work (Figure 2; Dataset a). The ER2 condition did demonstrate significant nuclear staining in the ES2 cell pellet with minimal background staining in the RMG-II cell pellet ( Figure 2; Dataset b). However, the staining in the ER1 condition was determined to give the best signal:noise ratio with no background cytoplasmic staining and crisp nuclear staining for the cell pellet (Figure 2; Dataset c). Control slides, omitting the primary antibody, were negative except for the ER2 condition in the RMG-II cell pellet where a weak cytoplasmic background could be seen (Figure 2; Dataset d). Thus there was minimal background inherent in the staining procedure. It was therefore determined that the antibody showed specificity for formalin-fixed paraffin embedded tissues and could be run on murine tissue.
Murine uterine tissue was used as positive control tissue samples, given the literature data on cell lines and endometrial tissue. The ER1 condition at 1 µg/ml demonstrated clean nuclear staining in the uterine epithelial compartment as well as nuclear staining of stromal cells. However, the nuclear staining in the stroma was not universal and distinct negative nuclei could be seen ( Figure 3A; Dataset e). There was no cytoplasmic or extracellular stromal background staining present and the antibody titrated successfully losing the intensity of staining, as expected (Dataset e). Following this, a concentration of 0.5 µg/ml was used for future preparations which provided clear and consistent staining in repeated batches using a different antibody lot (Dataset f). A No Primary antibody control (NPA) showed no staining in the epithelial or nuclear compartment ( Figure 3B; Dataset e).
Finally, when applied to a genetically engineered, tamoxifeninduced, Arid1a knockout mouse model, the staining in the uterine epithelium could be almost completely abrogated (Arrow, Figure 4b) when compared to the same area in a wild-type animal (Arrow, Figure 4a) with a small focal area of epithelial staining still present.
There was no effect of the KO on the staining in the stromal compartment.  Detailed legends for each dataset (Datasets a-f) can be found in the text file provided. For Version 2 an additional image was added to Dataset e.

Conclusions
It is clear from the use of ES2 and RMG-II cell lines that the Atlas Antibodies ARID1A antibody is specific for ARID1A in both Western blots and formalin-fixed paraffin embedded preparations of human origin and, coupled with the literature evidence, that it is validated in human tissue.
The staining pattern when applied to murine uterus showing a clear nuclear pattern, where there is a high level of sequence homology between the two species, is again consistent with the literature on this protein. When stained on an ARID1a knockout mouse model, the staining could be almost completely abrogated in the epithelial compartment but not in the stroma. Knockout mice generated in this manner are almost never 100% complete as in some cells recombination will not be induced due to issues such as low ligand (Tamoxifen) penetration or failure of the ligand to induce recombination, thus explaining the small focal area of epithelial staining. The difference in staining in the two compartments is also likely related to the same effect, as all other controls, such as omission of primary antibody remained negative. Thus, given the overwhelming data from other sources, it is likely the stromal staining reflects continuing Arid1a expression in this specific model system.
Therefore, in conclusion when taken in combination, it is clear that the anti-human ARID1a antibody is cross-reactive with murine tissue and can be used for this purpose.

Data availability
F1000Research: Dataset 1. Whole slide images from antibody validation of HPA005456 for immunohistochemistry -Version 2, 10.5256/f1000research.5514.d41579 10 Author contributions WH wrote and conceived the idea behind the article, IG performed the Western blots and mouse experiments and requested validation of ARID1A in murine tissue, JM performed all immunohistochemical staining.

Competing interests
No competing interests were disclosed. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

4.
This article, by Howatt validates an antibody to ARID1A in murine FFPE samples, is a good example et al. of the correct validations procedures that must be met when validating an antibody for use in tissue sections. Issues are relatively minor: What was the rationale for a starting concentration of 1μg/ml for initial validations?
There is a focal region of the KO epithelium which clearly appears to be positive. The authors would need to explain this.
Can the authors please add more discussion of the nature of the staining of the stromal cells. Do they consider this to be non-specific? Is such staining of stroma present in human tissue sections?
Control slides, omitting primary antibody, were carried out during the cell line phase of the experiments. Were such slides run on the murine tissue sections and can the results be included in the data. I note that the stromal staining was still present at selected concentration of 0.5μg/ml (dataset f) I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
No competing interests were disclosed. Thank you for your comments which we agree with and have taken on board. Specifically: All of the validations conducted at the Histopathology/ISH facility start with a dilution of 1:100 from the manufacturers provided stock concentration and we perform 3 retrieval methods; Sodium Citrate pH6, Tris EDTA pH9 & Proteinase K. We titrate as appropriate following establishment of the optimal retrieval method. The manufacturers original stock concentration was 100ug/ml, hence the 1ug/ml starting concentration. They have subsequently changed their stock concentration to 200ug/ml.
We believe that the knockout model recombining a floxed Arid1a with CRE-ERT2 is not 100% complete as in some cells recombination will not be induced due to issues such as low ligand (tamoxifen) penetration or failure of the ligand to induce recombination. Thus explaining the area of focal staining in the epithelium. Similarly, although we have not investigated this further, we believe that the same problem underlies the stromal staining and is likely due to penetrance of the tamoxifen after being delivered intraperitoneally. Thus, given the wealth of data and the no primary antibody controls, now included in Fig 3, we feel that the stromal staining is not non-specific. Stromal staining does indeed occur in human tissue. All of the above have now been included in the conclusion section of the article.
Kind Regards, , Cancer Research UK Cambridge Institute, UK Will Howat Dear Dr Chalmers, Many thanks for your comments which we agree with and have taken on board. Specifically: We have modified the title to read "Application of anti-ARID1a antibody..." We have BLAST searched the protein sequence and it confirms 95% homology for the sequence used for immunisations. This has been added to the methods.
We accept that the methods detailing the Western blotting are more limited than the IHC methods, as is the case of the methods detailing the production of the knockout mouse model. However, we feel that the article was designed as an example of antibody validation for immunohistochemistry and that while western blotting provides important additional data, it is not the focus of the article. We thus feel that there is sufficient information in the details behind the western blotting to repeat the experiment, without clouding the article with the full methods.
The sources of the cells and mouse tissue have now been cited.
The lack of staining of RMG-II cells was consistent with the literature (Anglesio ) and et al. this information has been added to the results section.
The figure legend has been modified.
We believe that the focal staining is a result of incomplete recombination floxed Arid1a allele and Cre/ERT2. The resulting knockout is almost never 100% complete as in some cells recombination will not be induced due to issues such as low ligand (tamoxifen) penetration or failure of the ligand to induce recombination. Similarly, we believe that the staining in the stromal compartment is specific but represents a failure of delivery of Tamoxifen or of recombination. The data from no primary antibody (NPA) control which are all completely clean, now included in Fig 3, helps to demonstrate this. We have included these points in the conclusion.
The author contributions have been modified. Kind Regards,

Will Howat
No competing interests were disclosed. Competing Interests: