The majority of β-catenin mutations in colorectal cancer is homozygous

To compare the findings of homozygous CTNNB1 in CRCs with those in extracolonic cancers we analysed panel sequenced lung cancers and melanomas. The Lung cancer cohort consisted of 23 non-small cell lung cancer analysed for EGFR mutations and the melanoma cohort of 10 metastasized melanomas analysed for BRAF mutations. Cancers were panel sequenced for therapy predictive purposes from January 2015 to December 2017 in the Department of Pathology, Charité University Medicine Berlin.

Biallelic mutations are defined by mutations in both, the maternal and paternal allele copy. Both copies may harbour the same mutation (homozygous), one copy may be mutated while the second copy is lost (hemizygous), or both copies may be affected by different mutations (compound heterozygous).

While detection of compound heterozygous mutations is straight forward, the interpretation of homo- or hemizyous mutations in cancer tissue is hampered by admixed non-neoplastic cells, which dilute tumorous DNA and thus lower the proportion of cancerous mutations. In routine practice, DNA isolated from cancers is “contaminated” by 20–70% of DNA from non-neoplastic cells. To score CTNNB1 mutations for mono- or biallelic mutations we followed expected mutational allele frequencies (AF) for tumor specific somatic mutations in samples with varying tumor cell purity as described previously [18].

In brief, we calculated the expected AFs of heterozygous and homo- or hemizygous mutations for different gene copy numbers and tumor cell purities. We defined mutations in genes typically mutated heterozygously in CRC as heterozygous reference mutations. For this purpose mutated AFs of oncogenes (KRAS, PIK3CA) where used as reference mutations for comparison with CTNNB1. In cases without mutations in oncogenes, the AF of mutated FBXW7 was evaluated. Despite being considered a tumor suppressor, FBXW7 mutations are almost always heterozygous in CRC [19].

We reasoned that irrespective of the tumor cell purity the AF of a homozygous mutation is twice as high as the AF of a heterozygous mutation. Accordingly, a CTNNB1 mutation was scored as biallelic/homozygous when its AF was twice as high as the AF of a reference gene (criterion for homozygosity). Heterozygous mutations were scored when the AF of the CTNNB1 was as high as that of the reference mutation.

In CTNNB1 mutated CRCs without reference mutations, we concluded that AFs ≥50% cannot result from heterozygous mutations. In these cases we considered homo- or hemizygous mutations. Biallelic mutations with retention of a wild type copy ^{(mut/mut/wildtype)} were considered unlikely due to the demonstration of wild type loss by RNA sequencing, the high tumour cell purity required to exceed an AF of 50%.

We therefore scored CTNNB1 mutations in CRCs without reference mutations as biallelic/homo- or hemizygous when the AF was ≥50% (criterion for homo-or hemizygous mutation) and as heterozygous when the AF of the CTNNB1 mutation was < 50%.

After microscopic identification of areas with highest tumor cell concentration, DNA was isolated using a commercial DNA Extraction Kit (DNeasy; Qiagen) following the manufacturer’s protocol.

In cohort 1, multiplex PCR-based amplicon high-throughput sequencing was performed using the Ion Torrent system (ThermoFisher Scientific) according to the manufacturer’s recommendations. Briefly, 10 ng of genomic DNA was used for library construction with the Ion AmpliSeq Colon and Lung Cancer Panel v2 and the Ion AmpliSeq Library Kit 2.0 (ThermoFisher). Library quantification was carried out on the StepOne Plus Real-Time PCR System employing the Ion Library TaqMan Quantitation Kit (both ThermoFisher). Emulsion PCR and subsequent enrichment was done on the Ion Chef instrument (ThermoFisher). Data analysis was performed using the Sequence Pilot software (JSI medical systems GmbH). Mean sequencing depth for CTNNB1 exon 3 and commonly heterozygous CRC mutations (KRAS, PIK3CA, FBXW7) was 1816 (range 569–4241) and 1308 (range 482–3671), respectively.

CRCs from cohort 2 were analysed with more extended customized panels as previously reported [17]. Mean sequencing depth for CTNNB1 exon 3 and commonly heterozygous CRC mutations (KRAS, PIK3CA, FBXW7) was 6144 (range 1891–11,334) and 3473 (range 1342–9467), respectively.

For single step generation of cDNA and PCR amplification we used the Invitrogen one step RT PCR KIT with platinum Taq (Fisher Scientific) according to the manufacturer’s protocol. PCR consisted of a single reverse transcription step at 55 °C for 30 min, followed by denaturation at 94 °C for 2 min and 40 cycles each consisting of a denaturation (94 °C for 15 s), annealing (60 °C for 30 s), and an extension step (68 °C for 1 min). PCR products were analysed by Sanger sequencing.

Colorectal cancer cell lines DLD1 were purchased from Horizon, HCT116 and LS180 from ATCC. The reported mutational status of CTNNB1 (wt/wt for DLD1, del45S/wt for HCT116, and S45F/S45F for LS180) (https://​www.​ncbi.​nlm.​nih.​gov/​pubmed/​24755471) was confirmed by DNA Sanger sequencing. DLD1, HCT116, and LS180 cells were cultured in RPMI medium, DMEM supplemented with 2 mM Ultraglutamine (Lonza), or EMEM, respectively, each containing 10% FCS and 100 U/ml penicillin-streptomycin (Biochrom).

After reaching 70–80% confluence, cells were harvested for immunofluorescence staining. Therefore, cells were washed with PBS twice and fixated with 4% formaldehyde for 30 min. Subsequently, cells were scratched from the plate using a spatula, and transferred in PBS into 15 ml Falcon tubes. Scratched cells in PBS were centrifuged for 2 min at 1200 rpm. The pellet was resuspended in histogel (ThermoScientific) at 70 C°, transferred into a histo cassette, embedded into paraffin, and cut in 3–4 μm sections. For deparaffinization of paraffin sections, slides were incubated in xylene for 5 min three times followed by rehydration in descending dilutions of ethanol in water (100, 90, 80, 70, 0%) for 5 min each at room temperature. For antigen unmasking, slides were incubated in citrate antigen retrieval buffer (82 mM trisodium citrate dihydrate, 18 mM citric acid monohydrate, adjust to pH 6.0) for 20 min at 100 °C, and unspecific binding sites were blocked by incubation with Peroxidase-Blocking Solution (Dako) for 30 min at room temperature. Slides were incubated with primary antibodies (Rabbit anti-β-catenin (#9562 Cell Signaling) 1:400, Mouse anti-E-cadherin (#14472 Cell Signaling) 1:200) overnight at 4 °C. For immunostaining, slides were subsequently incubated with secondary antibodies (Donkey anti-rabbit IgG Alexa Fluor 488, Goat anti-mouse IgG Alexa Fluor 555 (both Invitrogen), both 1:600) for 1 h at room temperature. Cell nuclei were visualized by incubation with 0.1 μg/ml DAPI in TBS. Slides were covered with Vectashield® mounting medium and coverslips. Images were acquired using an Axiovert flouresence microscope equipped with an Axiocam 506 color camera (both Zeiss). All incubation steps were followed by washing the slides with TBS three times. Antibodies were diluted in Antibody Diluent (Zytomed).

Protein was retrieved at 70–80% confluent growth. After washing twice with PBS, cells were scratched from the plates, transferred into a 1.5 ml tube, and centrifuged for 5 min at 5000 rpm at 4 °C. The pellet was resuspended in M-Per buffer (Thermo Scientific) supplemented with cOmplete protease inhibitor mix (Roche) and Phosphostopp phosphatase inhibitor mix (Roche). After sonification for 5 min and centrifugation at 13000 rpm for 10 min at 4 °C, protein concentration in the supernatant was measured using the BCA Protein Assay Kit (Pierce) according to the manufacturer’s protocol. Twenty microgram of protein in 6x loading buffer (300 mM Tris, 600 mM DDT, 12% SDS, 60% glycerol, bromphenol blue according to desired intensity) was heated to 95 °C for 10 min, cooled on ice and loaded on 2% SDS stacking/10% SDS separation gels. PageRuler Prestained Protein Ladder (Thermo Scientific) was used as protein weight standard.

Electrophoresis was run at 100 V. Subsequently, proteins were transferred to nitrocellulose membranes by semi-dry blotting between filter papers soaked with blotting buffer (25 mM Tris, 150 mM glycine, 10% methanol). Nitrocellulose membranes were blocked with 5% milk powder in TBST for 30 min. Incubation with primary antibodies (Rabbit anti-β-catenin (#9562 Cell Signaling), Mouse anti-E-cadherin (#14472 Cell Signaling), Rabbit anti-β-tubulin (#2146 Cell Signaling), all 1:1000) was carried out overnight at 4 °C, incubation with secondary antibodies (horse anti-mouse HRP-linked IgG; 7076S or goat anti-rabbit HRP-lined IgG 7074S, both Cell Signaling), was carried out for 1 h at room temperature. Protein bands were detected using the ECL Western Blot Detection Reagent (Amersham) according to the manufacturer’s protocol. All incubation steps were followed by washing the nitrocellulose membranes with TBST three times. Antibodies were diluted in 5% milk powder in TBST.

Formalin-fixed and paraffin-embedded tissue samples of CRC were cut into 4 μm sections. A BenchMark XT immunostainer (Ventana Medical Systems, Tucson, AZ) was used for subsequent immunohistochemical staining. For antigen retrieval, sections were incubated in CC1 mild buffer (Ventana Medical Systems, Tucson, AZ) for 30 min at 100 °C. Sections were stained with anti-E-cadherin antibody (clone 4A2C7, 1:50, Zytomed) for 60 min at room temperature, and visualized using the avidin-biotin complex method and DAB. Cell nuclei were counterstained by incubation with hematoxylin and bluing reagent (Ventana Medical Systems, Tucson, AZ) for 12 min each.

Tissue blocks for immunohistochemistry were available for 12 CTNNB1 mutated CRC and 5 CRC with CTNNB1 wild type. All 17 CRCs were MSI-H.

The stains were evaluated using an Olympus BC50 microscope and Olympus PLN 4X/0.1 Plan Achromat Objectives (Olympus Europe Holding GmbH, Hamburg, Germany). Histological images were acquired with the digital camera moticam 3.0 mp (Motic Instruments Inc., California, USA) and Motic Images Plus software (Motic instruments Inc.).

Among 839 CRCs from the diagnostic cohort we detected 16 cases with CTNNB1 mutations (1.9%), which comprised 7 MSI-H and 9 MSS tumors. The 7 CTNNB1 mutated MSI-H CRCs included 3 tumors analysed for LS diagnostics and one metastasized tumor from an LS patient analysed for companion diagnostics. LS status was unknown in the remaining 3 CTNNB1 mutated MSI-H CRC sequenced for therapy prediction. None of the CTNNB1 mutated MSI-H CRCs carried a BRAF mutation.

Among the total of 27 CTNNB1 mutated CRCs of both cohorts, CTNNB1 mutations in 13 CRCs (2 MSS, 11 MSI-H) were scored as biallelic/homozygous. Values for CTNNB1 AFs relative to the AFs of reference mutations were highly concordant with expected values and ranged between 1.9 and 2.1. Only CTNNB1 mutations in two CRCs displayed relative values of 2.4 and 3.1. Given a near perfect match of the relative AFs with that calculated for 3 and 4 mutated allele copies in the absence of wild type alleles, respectively, these tumours were also scored as homozygous.

In 10 out of 13 CRCs with homozygous CTNNB1 mutations the AFs of mutated CTNNB1 exceeded 50%.

The results are summarized in Table 1 and the supplementary Table 1.

The homozygous nature of CTNNB1 mutations was exemplarily confirmed by RNA sequencing of CTNNB1 in four cases, in which DNA sequencing revealed CTNNB1 mutations with AFs ranging from 50 to 63% (Fig. 1). In all four cases, wild type CTNNB1 RNA was absent or minimally present. Remnants of wild type CTNNB1 were considered to result from expression of CTNNB1 by stromal cells.

None of the 23 lung cancers with CTNNB1 mutations fulfilled the criterion for biallelic/ homozygous mutation and one tumor displayed a CTNNB1 AF of 57% and thus was scored as biallelic/homo- or hemizygous. The remaining 22 lung cancers and all 10 CTNNB1 mutated melanomas were scored as monoallelic/heterozygous.

In accordance with previous reports, we found the frequency of CTNNB1 mutations to be higher in MSI-H CRC than in MSS CRC and our data indicate that among the MSI-H CRC CTNNB1 mutations are mainly contributed by LS-associated tumors. We also discovered that not only the frequency, but also the type of CTNNB1 mutations differed between MSI-H and MSS CRCs. All MSI-H cancers displayed CTNNB1 mutations affecting either residue T41 or S45. The majority of alterations were missense mutations while two cases displayed 3 bp deletions at codon 45 (S45del) and one tumour harbored a 12 bp deletion (delS45-G48).

In MSS cancers, mutations were more evenly distributed among mutational hotspots. We detected in-frame deletions in 4 cases overlapping at codons 25–37, and 5 missense mutations at codon 34, 37, 41 (2), and 45 (Table 1, Fig. 2).

In order to gain an insight into the association of CTNNB1 mutations with E-cadherin expression, we studied β-catenin and E-cadherin levels in different MSI-H CRC cell lines by western blot analysis and immunofluorescence microscopy. Interestingly, we observed that β -catenin protein levels were similar in all three cell lines analyzed (DLD1: wildtype/wildtype, HCT116: wildtype/S45del, LS180: S45F/S45F). However, E-cadherin protein levels were significantly lower in LS180 than in DLD1 and HCT116 cells. This difference in E-cadherin protein levels could also be seen by immunohistochemistry. We further observed that LS180 cells displayed an incomplete, somewhat patchy E-cadherin distribution along the cell membrane while DLD1 and HCT116 cells showed a complete and distinct membrane staining of E-cadherin. Immunofluorescence with co-staining of β-catenin and E-cadherin confirmed the irregular distribution pattern E-cadherin in LS180 cells seen by immunohistochemistry. It further showed that this phenotype is associated with a diffuse cytoplasmic localization of β-catenin. In contrast, in DLD1 and HCT116 cells, β-catenin and E-cadherin are mainly co-localized along the cell membrane (Fig. 3) while only single cells in HCT116 displayed cytoplasmic and nuclear staining of β-catenin.

In vitro, we found that homozygous CTNNB1 mutation at residue S45 might perturb E-cadherin-mediated cell-cell adhesion. Hence, we questioned if CTNNB1 mutations found in CRCs also have an impact on E-cadherin expression.

Therefore, we analyzed the E-cadherin expression in 17 MSI-H CRCs with known CTNNB1 mutational status (10 homo- or hemizygous, 2 heterozygous, 5 wildtype) Strong reduction of E- Cadherin was found in 11 cancers while 6 CRCs displayed an expression similar to that in surrounding epithelium (Fig. 3). Aberrant expression of E-cadherin, however, was neither exclusive to CTNNB1-mutated tumours nor did all CTNNB1 mutated CRCs display reduced E-cadherin expression. Three out of 5 CRCs (60%) without CTNNB1 mutations and 8 of 12 CRCs (66%) with CTNNB1 mutations showed a reduced concentration of membranous E-cadherin while normal E-cadherin expression was found in 2 CRCs without and 4 CRCs with CTNNB1 mutations.