Relationship of microsatellite instability to mismatch repair deficiency in malignant tumors of dogs

2.1 Clinical samples

Clinical samples were obtained from tumor-bearing dogs that came to Yamaguchi University Animal Medical Center and Kyoto Animal Medical Center for diagnosis or treatment from May 2020 to March 2021. All dogs underwent biopsy or surgical resection and were diagnosed with malignant tumors by veterinary pathologists at Yamaguchi University or IDEXX Laboratories (Tokyo, Japan), except for some cases with lymphoma that were diagnosed only by cytology. Blood and fresh tumor tissues were collected from each dog to evaluate MSI, and an oral mucosal swab was taken instead of blood to extract normal tissue DNA when tumor cells were observed on a blood smear. Formalin-fixed paraffin-embedded (FFPE) tumor tissues were collected from each dog to evaluate dMMR. The study was approved by the Ethics Committee of Yamaguchi University.

2.2 PCR for microsatellite sequences

Genomic DNA was extracted from blood (or oral mucosal swab) and fresh tumor tissues with a QIAamp DNA Blood Mini Kit (QIAGEN Tokyo, Tokyo, Japan) or QIAamp DNA Mini Kit (QIAGEN Tokyo), respectively, and was stored at 4°C until further use.

For MSI analysis, 12 MS loci were selected (Table 1)^{24, 25} because of their high instability rates in a previous report³ and our preliminary experiments. Microsatellite sequences were amplified from tumor and normal DNA by PCR in 15 μL of reaction volume, including 12 ng of DNA and 2.5 μM of each primer labeled with fluorescent dye using BIOTAQ DNA Polymerase (Meridian Life Science, Inc, Memphis, Tennessee). The PCR amplicons were detected by capillary electrophoresis (3500xL Genetic Analyzer, Thermo Fisher Scientific K.K., Tokyo, Japan) and analyzed by a software application (Peak Scanner, Thermo Fisher Scientific K.K.). In each MS sequence, electropherograms of tumor and normal DNA were compared and defined as MSI-positive if a difference in product size was detected. In addition, LOH was defined as a ≥35% reduction in the peak of 1 allele in tumor DNA compared to that in normal DNA.²⁶ The electropherograms of each sample were evaluated independently by 3 observers.

2.3 Immunohistochemistry for dMMR analysis

As described in a previous report,⁴ each FFPE tissue was incubated with mouse monoclonal anti-human MLH1 antibody (clone G168-15, GeneTex, Inc, Irvine, California, diluted at 1:50), rabbit monoclonal anti-human MSH2 antibody (clone SP46, Abcam, Cambridge, UK, diluted at 1:5100), or mouse monoclonal anti-human MSH6 antibody (clone 44/MSH6, BD Transduction Laboratories, San Diego, California, diluted at 1:250). The isotype antibodies, rabbit IgG antibody (clone DA1E, Cell Signaling Technology, Inc, Danvers, Massachusetts) and mouse IgG1 antibody (clone P3.6.2.8.1, eBioscience, Inc, San Diego, California), were used as negative immunostaining controls. A peroxidase stain DAB Kit (Nacalai Tesque, Inc, Kyoto, Japan) was used to visualize the immunoreaction. The sections were counterstained with Mayer's hematoxylin. To bleach the melanocytic granules of melanoma tissues, the sections were incubated in 10% hydrogen peroxide before antigen retrieval. A Simple Stain AEC Solution (Nichirei Bioscience, Inc, Tokyo, Japan) was used to visualize the immunoreaction instead of DAB.

Immunoreactivity was evaluated if the nuclei of the tumor cells were immunostained with each antibody as compared with those in isotype controls. The immunostaining of the normal tissue areas of each sample was used as a control and simultaneously evaluated to confirm positive immunostaining. Negative immunostaining was confirmed when the nuclei of tumor cells showed negative immunostaining. The immunoreactivity of each sample was independently evaluated by 4 observers. Based on the results of IHC, the MMR status of each case was determined; MMR proficiency (pMMR) was confirmed if all 3 MMR proteins showed positive immunostaining, or dMMR was confirmed if ≥1 proteins showed negative immunostaining based on criteria used in studies of human tissues.¹⁵

2.4 Cell culture

Human embryonic kidney (HEK) 293 T-cell line and SV40-immortalizated canine gallbladder epithelial cells (kindly provided by Dr. Kenji Baba) were cultured in a Dulbecco's modified Eagle medium (DMEM)-based complete medium (DMEM mixed with 10% fetal bovine serum and 100 units/mL penicillin-streptomycin and 55 μM 2-mercaptoethanol). All cell lines were maintained at 37°C in a humidified CO₂ gas incubator.

2.5 Establishment of dMMR/MSI cell lines

According to previous reports,²⁷ CRISPR-Cas9 small guide RNAs (sgRNAs), which target 1 or more of the MMR genes (mlh1, msh2, and msh6), were designed by an online CRISPR design tool (http://zlab.bio/guide-design-resources). The sequences of sgRNAs of canine MLH1, canine MSH2, and canine MSH6 were 5′-GTGGTGAACCGCATCGCGGC, GCGCTTCTTCCAGGCCATGC-3′, and 5′-GAAGTAAGGCCTAAGGTCCA-3′, respectively. These oligonucleotides were synthesized and inserted into a lentiCRISPR v2 plasmid, which was donated by Dr. Feng Zhang (Addgene, Cambridge, Massachusetts, #52961; http://n2t.net/addgene:52961; RRID: Addgene 52961). To produce lentivirus-expressing sgRNAs that target each gene, HEK293T cells were seeded at 3.0 × 10⁵ cells/well on a 6-well plate a day before transfection. On the day of transfection, 1.25 μg of each plasmid was mixed with PEI Max (Polysciences, Inc, Warrington, Pennsylvania) in an OPTI-MEM buffer (Thermo Fisher Scientific K.K.), incubated for 15 minutes at room temperature, and added onto HEK293T cells. After 24 hours, the supernatant of HEK293T cells was replaced with a DMEM-based complete medium without antibiotics. After another 24 hours, the supernatant containing lentiviruses was collected and transferred onto gallbladder epithelial cells. After 12 hours of infection, the supernatant was replaced with a DMEM-based complete medium, followed by the addition of 0.2 mg/mL hygromycin onto the gallbladder epithelial cells to select stably transduced cells. Individual single clones were isolated by limited dilution method and further analyzed by Western blotting to confirm the knockout of each protein. Based on the results of Western blotting, clones with complete knockout of each MMR gene were selected and passaged >30 times over 3 months after single clone isolation.

2.6 Western blotting analysis

Each cell line was collected and lysed with an NP-40 lysis buffer (1% NP40, 10 mM Tris HCl [pH 7.5], 150 mM NaCl, 1 mM EDTA). Equal amounts of protein samples (30 μg) were electrophoresed in 7% acrylamide gel before transferring them onto membranes. The membranes were incubated with each primary antibody (same as the antibodies used in IHC) diluted at 1:1000 in 0.5% skimmed milk in Tris Buffered Saline with Tween20 overnight at 4°C. Horseradish peroxidase (HRP)-conjugated anti-mouse IgG antibody (Merck Millipore, Darmstadt, Germany) or HRP-conjugated anti-rabbit IgG antibody (Jackson ImmunoResearch, West Grove, Pennsylvania) was used as a secondary antibody. The membranes were soaked in Western Lightning Plus-ECL (PerkinElmer, Tokyo, Japan) to visualize the immunoreaction. Immunoreactive bands were quantified and analyzed using AMERSHAM Image Quant 800 (Global Life Sciences Technologies Japan, Tokyo, Japan). The membranes were reused with mouse monoclonal anti-beta actin antibody (Sigma-Aldrich Japan K.K., Tokyo, Japan) to verify equal loading of proteins in each lane.

2.7 Statistical analysis

Statistical analysis was performed by JMP Pro 15 software (JMP Japan, Tokyo, Japan). A Wilcoxon rank-sum test was conducted to determine the association between the number of MSI-positive markers and dMMR/pMMR status. Statistical significance was set at P < .05.

A least absolute shrinkage and selection operator (LASSO) regression model and decision tree model were fitted to further select important variables for the association between dMMR/pMMR status (dependent variable) and MSI-positive markers (independent variables). The LASSO regression model shrank all regression coefficients toward zero and set the coefficients of irrelevant variables to zero according to the regulation weight λ. The optimal value of λ was determined by a 5-fold cross validation. The decision tree analysis used the same variables as LASSO (dMMR/pMMR status as dependent variable and MSI-positive markers as independent variable). The tuning parameters were determined by a 5-fold cross validation. Ten cases with other tumors (Table 2) were excluded in the LASSO regression and the decision tree model analyses.

3.1 Study population

In our study, 101 dogs with tumors were included. The characteristics of these dogs are shown in Table S1, Supporting Information. The population consisted of 51 males (27 castrated) and 50 females (38 spayed). Median age at diagnosis was 11 years and 11 months (range, 5-16 years and 10 months). The cases were classified into 25 cases of oral malignant melanoma, 13 cases of lymphoma, 3 cases of mast cell tumor, 4 cases of mammary gland carcinoma, 5 cases of urothelial carcinoma, 9 cases of hepatocellular carcinoma, 9 cases of squamous cell carcinoma, 7 cases of adenocarcinoma, 3 cases of hemangiosarcoma, 11 cases of soft tissue sarcoma, and 12 cases of other tumors (Table 2). Four lymphoma cases (cases 27, 29, 37, and 38 in Table S1) and 1 urothelial carcinoma case (case 46 in Table S1) had been treated before the tissues were collected for the study.

3.2 Microsatellite analysis for clinical cases

The results of MSI and LOH are shown in Table S1. Microsatellite instability was defined as an alteration in the peak size of MS amplicons in tumor DNA compared to that in normal DNA (Figure 1). At least 1 MSI-positive marker was detected in 20 cases of oral malignant melanoma, 5 cases of lymphoma, 4 cases of soft tissue sarcoma, 6 cases of hepatocellular carcinoma, 5 cases of squamous cell carcinoma, 3 cases of adenocarcinoma, 2 cases of urothelial carcinoma, 4 cases of mammary gland carcinoma, 2 cases of hemangiosarcoma, and 4 cases of others, whereas no MSI-positive marker was detected in mast cell tumors (Figure 2). Oral malignant melanomas tended to have MSI markers more frequently.

Loss of heterozygosity was defined as a reduction in the peak of 1 allele in tumor DNA compared to that in normal DNA (Figure 1) and was evaluated for all MS loci. In addition, some cases showed increased amplitude of the other allele, which could indicate copy neutral LOH (CN-LOH). Copy loss LOH (no difference in amplitude of the other allele) was observed frequently in markers located in chromosome of Canis lupus familiaris (CFA) 5 (FH3837, FH2594, and FH3113), CFA 16 (FH2175), and CFA 22 (C22.763) across tumor types (Figure 3 and Table S1). Multiple MS markers on CFA 5 showed copy loss LOH in 4 cases (cases 22, 78, 89, and 93 in Table S1).

3.3 Relationship between dMMR and MSI in clinical cases

The FFPE specimens of tumor tissues were obtained from 81 of 101 dogs. The canine MLH1, MSH2, and MSH6 (referred to cMLH1, cMSH2, and cMSH6) in these specimens were analyzed by IHC. The classifications of the 81 tumor specimens are shown in Table 3.

Representative results of IHC are shown in Figure 4A,B. Based on criteria for dMMR used in humans, all analyzed cases were classified into pMMR or dMMR. The proportion of tumors with dMMR was 8/21 (38.1%) cases of oral malignant melanoma, 1/4 (25%) cases of mammary gland carcinoma, 1/5 (20%) cases of urothelial carcinoma, and 3/9 (33.3%) cases of hepatocellular carcinoma, whereas the remainder of the specimens were classified into pMMR (Table 3 and Table S1). For 20 cases, including 2 dMMR cases (cases 1 and 46), out of 101 cases, FFPE samples of normal tissues also were available, and expression of MMR proteins was detected in all of them (Figure S1).

Then, whether or not dMMR was associated with MSI in clinical cases was confirmed. Comparing the number of MSI-positive markers between cases with dMMR or pMMR in entire tumors, these markers were significantly higher in cases with dMMR (Figure 4C, P < .0001). Further analysis of the relationship by tumor type identified significant results in oral malignant melanoma and hepatocellular carcinoma (P < .01 and P < .02, respectively) and entire tumors, but not in urothelial carcinoma and mammary gland carcinoma (P = .11 and P = .08, respectively). On the other hand, there was no relationship between dMMR and number of LOH (data not shown).

3.4 Statistical evaluation to select MS loci suitable for dMMR/pMMR status

To identify which MSIs contributed substantially to dMMR in tumors of dogs, the LASSO regression model and decision tree analysis were used. In LASSO, 6 variables were estimated as nonzero coefficients (but not statistically significant), which might indicate that the appearance of MSI in FH2305 and the total number of MSI-positive markers reflect dMMR (Table 4).

Additionally, the decision tree model indicated that the cases with dMMR or pMMR could be classified with high probability by setting up nodes with ≥2 or ≤1 MSI-positive loci, respectively (Figure 5).

3.5 Establishment of cell lines with dMMR and MSI analysis

We investigated whether dMMR induces MSI in vitro by canine cell lines after knockout of the MMR gene. A CRISPR/Cas9 system was used to establish the dMMR cell lines, and was followed by confirmation of knockout by Western blotting (Figure 6) and immunocytochemistry (Figure S2). No MLH1 protein was detected in both clones of the MLH1 knockout cell line. The MSH2 and MSH6 knockout cell lines expressed no MSH6 and MSH2, in addition to defective expression of MSH2 and MSH6, respectively, as shown in a previous study.⁴ The knocked-out cell lines and wild-type cells were serially passaged for 3 months to allow mutations to accumulate in MS loci. Then, the MSI status in these cell lines was analyzed at 2 time points (0 and 3 months). Based on results of the MSI analysis, an MSI-positive marker occurred in dMMR cell lines but not in wild-type cell lines from 0 to 3 months. At 3 months, 2 loci (FH2054 and AHT137) showed MSI in a cMLH1-deficient cell line, 2 loci (PEZ11 and FH3837) in a cMSH2-deficient cell line, and 4 loci (PEZ11, FH2054, FH2594 and AHT137) in a cMSH6-deficient cell line (Table 5).