Gene fusions and oncogenic mutations in MLH1 deficient and BRAFV600E wild-type colorectal cancers

Universal dMMR screening for all newly diagnosed CRC cases has been routine in Helsinki University Hospital (HUH) since January 2018, using IHC for MMR proteins MLH1, MSH2, MSH6 and PMS2. In case of deficient MLH1 IHC, the dMMR screening algorithm leads to BRAFV600E mutation–targeted IHC testing to identify potential LS patients, i.e. immunonegative and thus BRAFV600Ewt, to be further tested for MLH1ph (methylation-positive cases interpreted as sporadic dMLH1). Our patient material consisted of consecutive CRC patients undergoing primary surgery at HUH between January 2018 and April 2020 (n = 2079), out of which 66 showed immunohistochemically confirmed loss of expression of MLH1 (and concomitant loss of PMS2) and BRAFV600Ewt status. Two of the 66 samples were not representative for our study and two were not available leading to a study cohort of 62 samples. The study was approved by the Ethics Committee of the HUH.

MLH1 methylation status of the 62 CRC cases was determined with methylation-specific multiplex ligation-dependent probe amplification (MS-MLPA). Here, SALSA MS-MLPA Probemix ME011-D1 Mismatch Repair Genes kit (MRC Holland, Amsterdam, the Netherlands) was used to determine the methylation status of the promoter regions of MMR genes (MLH1, MSH2, PMS2 and MSH6) using probes that contain a digestion site for the methylation-sensitive restriction endonuclease HhaI. At the same time, the BRAFV600E point mutation was detected with a probe specific for this mutation. The kit includes six probe pairs for MLH1 (covering A to D sites in MLH1 promoter and MLH1 intron 1 93 nt after exon 1). All reactions and analysis of the results were accomplished according to the manufacturer’s instructions and as described in Gylling et al. [9]. DNA amount of 250 ng extracted from formalin-fixed paraffin-embedded (FFPE) tissue samples was used for each MLPA reaction. As a threshold, ratio of > 0.15 (corresponding to > 15% of methylated DNA) was used to indicate promoter methylation. In addition, 45/62 CRC cases were analysed using bisulphite pyrosequencing of the MLH1 promoter region to detect methylated cytosines in tumour DNA at positions c.-269, c.-262, c.-252 and c.-250 of the MLH1 gene, showing 100% concordant results.

KRAS and NRAS mutations were identified, and the BRAFV600wt status confirmed, from the 62 CRC FFPE samples using real-time PCR–based CE-IVD–validated Idylla KRAS and NRAS-BRAF mutation assays (Biocartis NV, Mechelen, Belgium). Furthermore, the MSI status of the dMMR samples was confirmed with Idylla MSI test that was performed as described previously [10]. The Idylla KRAS Mutation Test detects 21 KRAS mutations in exons 2, 3 and 4, whereas the Idylla NRAS-BRAF Mutation Test detects 18 NRAS mutations in exons 2, 3 and 4, and five BRAF mutations in codon 600. For the analyses, 10-µm thick tissue slices were cut from the CRC tissue blocks. The tissue sample handling and analysis were performed according to the manufacturer’s protocol and 20 to 80% of tumour cell content was used.

The cohort of 62 dMLH1/BRAFV600Ewt CRC FFPE samples underwent RNA–based targeted NGS analysis using novel RUO FusionPlex® Lung v2 (Invitae Corporation, San Francisco, CA) which identifies fusion transcripts of ALK, BRAF, EGFR, ERBB2, FGFR1, FGFR2, FGFR3, KRAS, MET, NRG1, NTRK1, NTRK2, NTRK3, NUTM1, PIK3CA, RET and ROS1 genes. Total nucleic acid was extracted from 10-µm FFPE tissue sections using QIASymphony RNA Kit (QIAGEN, Valencia, CA), and the quality was assessed by Qubit RNA HS Assay Kit (Thermo Fisher Scientific, Waltham, MA). RNA was then reversely transcribed, and the quality was checked by ABI StepOne Plus (Applied Biosystems/Thermo Fisher Scientific). Libraries were quantified with KAPA Library Quantification Kit Illumina® Platforms (ABI StepOne Plus). Libraries were paired-end sequenced at 2 × 150 cycles on Illumina Novaseq 6000 instrument using SP flow cell with Novaseq Xp workflow (Individual Lane loading). Data were analysed using the Archer Analysis v6.2.7 software (Archer/Invitae) for the presence of gene fusion using GRCh37 as the reference genome (Supplementary Table S1).

NTRK1 fusions were further validated with two independent RNA–based NGS platforms, FusionPlex Comprehensive Thyroid and Lung panel (CTL AK0070 v1.1, Invitae) and TruSight RNA Pan-Cancer panel (Illumina, San Diego, CA). FusionPlex CTL Kit detects fusion transcripts of ALK, AXL, BRAF, CCND1, FGFR1, FGFR2, FGFR3, MET, NRG1, NTRK1, NTRK2, NTRK3, PPARG, RAF1, RET, ROS1 and THADA genes. In short, total nucleic acid was extracted from 10-µm FFPE tissue sections using QIASymphony RNA Kit (QIAGEN), and the quality was assessed by Qubit RNA HS Assay Kit (Thermo Fisher Scientific). RNA was then reversely transcribed, and the quality was checked by ABI StepOne Plus (Applied Biosystems/Thermo Fisher Scientific). Libraries were quantified with Ion Library TaqMan Quantitation Kit (ABI StepOne Plus) and sequenced on either Ion S5 or Ion Proton systems (IonTorrent/Thermo Fisher Scientific). Data were analysed using the Archer analysis software (Suite_Analysis_v6.0.4; Invitae) for the presence of gene fusion using GRCh37 as the reference genome (Supplementary Table S1).

For Illumina TruSight RNA Pan-Cancer panel, total RNA was extracted from 10-µm FFPE tissue sections using RNeasy FFPE kit (QIAGEN) according to the manufacturer’s protocol. The RNA quality (DV200) was assessed using the Agilent Bioanalyzer 2100 instrument (Agilent, Santa Clara, CA). Targeted RNA libraries were prepared according to the Illumina TruSight RNA Pan-Cancer panel reference guide (Illumina), and the amount of input RNA was evaluated based on DV200 value. The final libraries were paired-end sequenced at 2 × 75 cycles on Illumina MiniSeq instrument using its High Output kit. Illumina DRAGEN software (v3.8.4) was used for the fusion calling and GRCh38 was used as the reference genome (Supplementary Table S1).

NGS analysis of oncogenic mutations in a metastasized case (#52) with synchronous CRC was done using an in-house cancer panel containing seven target genes (PIK3CA, EGFR, KIT, KRAS, MET, NRAS and PDGFRA) and exons 11–15 of BRAF and was performed as previously described [11].

All 62 cases were further tested with the RUO real-time RT-PCR–based Idylla™ GeneFusion Assay (Biocartis NV) to detect ALK, ROS1, RET and MET exon 14 skipping and NTRK1/2/3 fusions in single cartridge and approximately in 3 h. For NTRK1/2/3, ALK, ROS and RET fusions, the detection was performed with expression imbalances indicating putative gene fusion irrespective of fusion partner based on the 3’ kinase overexpression caused by the partner gene. For ALK, ROS and RET fusions and MET ex14 skipping, the detection was additionally performed by assessing specific gene fusion variations of the most common variants. For the analysis, 10-µm sections were cut from the FFPE CRC tissue blocks. The tissue sample handling and analysis were performed according to the manufacturer’s protocol and 20 to 80% of tumour cell content was used.

For immunohistochemical analyses, 4-µm sections cut from the FFPE CRC tissue blocks (n = 62) were used. The MMR IHC was performed as described previously [10] and BRAFV600E mutation was detected by using the specific monoclonal ready-to-use antibody (clone VE1, 760–5095, Roche, Tucson, AZ), utilising detection with OptiView DAB kit (760–700, Roche) and additional Amplification Kit (760–099, Roche) [12]. Pan-TRK IHC was performed using ready-to-use monoclonal antibody (clone EPR17341, 790–7026, Roche) and detection by OptiView DAB kit (760–700, Roche). ALK IHC was performed using monoclonal antibody (clone 5A4, Novocastra™, Leica Biosystems, Newcastle Upon Tyne, UK) at dilution 1:50 and Roche’s OptiView DAB and Amplification kits. All above-described immunostainings were performed with the Ventana Benchmark ULTRA immunostainer (Roche). The other pan-TRK antibody used in this study (clone A7H6R, #92,991, Cell Signaling Technology Inc., Danvers, MA) was diluted 1:50 and stained with Autostainer (Agilent) using Envision Flex High pH detection kit (K8000, Agilent). Each slide had a non-neoplastic colorectal specimen as a positive external control (neural structures), and since we used (freshly cut) whole tissue sections of the tumour samples, most of them also contained positive internal control (neural structures in the muscularis propria). IHC stainings were analysed in a blinded manner by IU and AR. Positive pan-TRK immunoreactivity, indicating putative NTRK fusion, was determined when ≥ 1% of the cancer cells displayed cytoplasmic, membranous, nuclear and/or perinuclear immunopositivity.

The RNA–based NGS panel (FusionPlex Lung v2) was considered as the gold standard test against which the overall agreement, sensitivity and specificity, and the positive and negative predictive values (PPV and NPV, respectively) were calculated. Fisher’s exact test (GraphPad QuickCalcs: https://​www.​graphpad.​com/​quickcalcs/​contingency2/​) was used for comparison of the agreement between NGS and the Idylla test, and between NGS and the IHC methods. Clinicopathological characteristics between fusion positive and negative cases were compared using Mann–Whitney U-test (numerical variables) or Fisher’s exact test (categorical variables). All statistical tests were two-tailed and numeric variables are reported by median and range. P-value less than 0.05 was considered as statistically significant.

Our study consisted of 62 CRC samples with dMLH1 and BRAFV600Ewt status as detected by IHC. Of these patients, 59.7% were females and the tumours localised mainly to the right colon (85.5%), were low-grade (69.4%), pT3 (59.7%), pN0 (64.5%) and M0 (80.6%) (Table 1). Of the cases, 42/60 were MLH1ph as detected by MS-MLPA. One case was not available for the analysis. One case was not analysable despite three attempts, but was successfully analysed using bisulphite pyrosequencing and found to be MLH1ph. Overall, 43/61 (70.5%) of the CRC cases were MLH1ph (Supplementary Table S2). KRAS mutations were detected in 20 cases by Idylla KRAS mutation test, and three NRAS mutations were detected by Idylla NRAS-BRAF mutation test (Supplementary Table S2). Overall, RAS mutations were found in 23 (37.1%) of the dMLH1/BRAFV600Ewt CRC cases. KRAS mutations were the most common in codon 12 (10 cases), followed by codon 61 (five), codon 146 (four) and codon 13 (one). NRAS mutations were found in codons 61 (two cases) and 12 (one case). MSI status was verified using Idylla MSI test, which resulted MSI in 62/62 of the cases (one sample, case 16, was initially MSS but after re-analysis using a more representative tissue block resulted MSI). Finally, the NRAS-BRAF test (n = 62) and the MS-MLPA (n = 60) analyses confirmed the BRAFV600Ewt status of all CRC cases.

Expanded RNA–based NGS panel FusionPlex Lung v2 was used as the gold standard for detection of gene fusions in the 62 dMLH1/BRAFV600Ewt CRC samples. Seven in-frame NTRK1 fusions (Table 2) were identified (TPM3::NTRK1, PLEKHA6::NTRK1 and LMNA::NTRK1, each in two cases, and IRF2BP2::NTRK1 in one case), of which IRF2BP2::NTRK1 is a novel fusion in CRC (Fig. 1). FusionPlex CTL and/or TruSight RNA Pan-Cancer NGS panels confirmed the seven NTRK1 fusions (Supplementary Table S3). An NTRK1 fusion was thus identified in 7/62 (11.3%) of the dMLH1/BRAFV600Ewt, in 7/43 (16.3%) of the dMLH1/BRAFV600Ewt/MLH1ph and in 7/30 (23.3%) of the dMLH1/BRAFV600Ewt/MLH1ph/RASwt CRC samples. In addition to NTRK1 fusions, Lung v2 NGS detected two in-frame ALK fusions (EML4::ALK in both), four in-frame RET fusions (CCDC6::RET in three cases and NCOA4::RET in one case) and seven in-frame BRAF fusions (AGAP3::BRAF and TRIM24::BRAF both in two cases, and STARD3NL::BRAF, MKRN1::BRAF and LMTK2::BRAF each in one case) (Table 2). Analysis by the Lung v2 NGS failed in four cases (#7, 9, 11 and 26). These four samples were reanalysed using FusionPlex CTL NGS kit, and three were found to be fusion negative, whereas one failed with this panel as well (case 7). A kinase fusion was thus identified in 20/62 (32.3%) of the dMLH1/BRAFV600Ewt, in 20/43 (46.5%) of the dMLH1/BRAFV600Ewt/MLH1ph and in 20/30 (66.7%) of the dMLH1/BRAFV600Ewt/MLH1ph/RASwt CRC samples.

Fusion-positive cases (n = 20) had significantly higher preponderance of MLH1ph (P = 0.0002) and RASwt genotype (P < 0.0001) compared to fusion negative cases (n = 42), but no statistically significant differences in clinicopathological features including age, sex, pTNM stage, grade or tumour site were observed between the negative and positive cases (Supplementary Table S4). One NTRK1, one ALK and one RET fusion case presented distant metastases (cases 12, 35 and 43) and were all RASwt. Furthermore, one NTRK1-rearranged case with distant metastases (case 52) had synchronous sigma tumour with proficient (p)MMR and KRASG12V mutation, as was also shown to be the case for the peritoneal metastasis, whilst the NTRK1-rearranged tumour in colon ascendens was dMMR and RASwt.

All 62 CRC samples with dMLH1 and BRAFV600Ewt status were next analysed using Idylla gene fusion assay. NTRK1 expression imbalance was identified in seven cases, which coincided with the Lung v2 NGS fusion cases (Table 2 and Supplementary Table S2). The concordance between Idylla gene fusion test NTRK1 expression imbalances and NTRK1 fusions detected by Lung v2 NGS was 100% (59/59 valid Idylla tests, P < 0.0001; 3/62 Idylla analyses were invalid). Using the NGS as a reference, the sensitivity, specificity, PPV and NPV of Idylla gene fusion test to identify expression imbalance in NTRK1 fusion cases were 100% for each. Idylla detected the Lung v2 NGS identified EML4(e6)::ALK(e20) fusion as a specific ALK fusion. The second ALK fusion identified by the Lung v2 NGS (EML4(e21)::ALK(e20)) is not included in the set of specific fusions detected by Idylla, and it was thus detected as an ALK expression imbalance. However, Idylla reported seven additional ALK expression imbalances in cases, which were not ALK fusion–positive by Lung v2 NGS. Thus, concordance between Idylla ALK specific fusion or expression imbalance and the Lung v2 NGS was 88.1% (52/59 valid Idylla tests, P = 0.021), and the sensitivity, specificity, PPV and NPV were 100%, 87.7%, 22.2% and 100%, respectively. In addition, two combined RET–specific fusion and RET expression imbalance results were detected in line with the Lung v2 NGS results, whilst two RET fusions were not detected by the Idylla analysis (specific fusion detection not included and no RET expression imbalance reported). Further, one false positive RET expression imbalance was reported. Thus, concordance between Idylla RET–specific fusion and/or expression imbalance and the Lung v2 NGS was 94.9% (56/59 valid Idylla tests, P = 0.010), and the sensitivity, specificity, PPV and NPV were 50%, 98.2%, 66.7% and 96.4%, respectively. Initially, seven Idylla results were invalid due to improper/inappropriate RNA amplification, of which four cases were successfully re-analysed (Table 2 and Supplementary Table S2). It should be pointed out that none of the initially invalid samples in the Idylla analyses was the same as the four samples that failed with the Lung v2 NGS. In addition, some individual targets remained invalid in the Idylla analyses despite re-runs, which was especially common in the expression imbalance of NTRK3 (12/59, 20.3%; Supplementary Table S2).

Pan-TRK IHC was used to screen for NTRK gene fusions in the 62 CRC samples using two antibody clones and scored in blinded manner by two observers independently. With clone EPR17341 eight and with clone A7H6R seven, CRC samples were scored as immunopositive by both observers. However, staining intensity was higher and subcellular localization clearer using the clone EPR17341 protocol and scoring results using this clone are shown in Table 2. All samples showed cytoplasmic immunopositivity with variable intensity that was relatively diffuse (50–100% of the tumour cells being positive), and four samples also expressed membranous and two perinuclear staining (Table 2, Fig. 2 and Supplementary Fig. S1). No nuclear staining was observed. The overall agreement between pan-TRK IHC (EPR17341) and NGS was 98.4% (61/62, P < 0.0001). Using the NGS as a reference, the sensitivity of pan-TRK (EPR17341) IHC was 100% and the specificity 98.2% with PPV of 87.5% and NPV of 100%. For the clone A7H6R, these parameters were 100%. Importantly, one EPR17341 pan-TRK IHC positive sample, expressing the lowest cytoplasmic staining intensity and being negative with clone A7H6R, did not contain an NTRK fusion but an ALK fusion (case 12) (Table 2 and Supplementary Figs. S1 and S2). This case, and the other ALK fusion case, showed strong cytoplasmic ALK immunopositivity, whilst the rest of the samples were completely negative (Table 2 and Supplementary Fig. S2).