Comparison of NTRK fusion detection methods in microsatellite-instability-high metastatic colorectal cancer

Patient selection and FFPE collection

All patients diagnosed with MSI-H/dMMR mCRC between 2015 and 2021 in the Netherlands were selected using the Netherlands Cancer Registry. All available formalin-fixed, paraffin-embedded (FFPE) tumor tissue blocks were retrieved using the Nationwide Network and Registry of Histo- and Cytopathology in the Netherlands (PALGA). Whenever possible, tissue of the primary tumor resection was used. If not available, tissue from a primary tumor biopsy or metastasis was used.

Immunohistochemistry screening

We used pan-TRK IHC as a screening test. The received tissue blocks were cut into 4-µm-thick slides and stained using a Ventana bench mark ultra autostainer using antigen retrieval for 24 min with CC1 (EDTA) and 32 min antibody exposure with a rabbit pan-TRK monoclonal antibody (mAb) (clone EPR17341, Abcam, Cambridge, MA) in a dilution of 1:500 according to standard procedures. Brain tissue was used as positive control.

All slides were examined by two independent qualified pathologists (MML and LAAB) and scored for percentage of positive tumor cells, intensity, and staining pattern (cytoplasmic, nuclear, perinuclear, and/or membranous). In case of discordancy between the pathologists, the highest intensity score was used underlining the use as screening method. Positive staining was defined as ≥ 1% of tumor cells showing staining above background in any pattern and any intensity. Tumors that showed weak, non-specific granular cytoplasmic staining (Fig. 1B) were scored negative. All tissue samples with positive staining and 11 negative samples with weak, non-specific staining were included in the comparative analysis and analyzed with RNA-NGS, FFPE-TLC, FISH, and qRT-PCR.

Fig. 1figure 1

Pan-Trk IHC, 40 × original magnification. (A) Negative staining, no additional testing was performed. (B) Weak, non-specific granular cytoplasmic staining (scored negative), no NTRK fusion was found. (C) Moderate cytoplasmic staining, no NTRK fusion was found. (D) Strong cytoplasmic + nuclear staining, LMNA::NTRK1 fusion was detected. (E) Strong cytoplasmic + perinuclear staining, LMNA::NTRK1 fusion was detected. (F) Strong cytoplasm + membranous staining, TPM3::NTRK1 was detected

In addition, a subset of pan-TRK negative cases was selected to examine the sensitivity of pan-TRK IHC. For this analysis, we selected the subgroup with the highest prevalence of NTRK fusions (MSI-H/dMMR, BRAF wild-type, and RAS wild-type) for FFPE-TLC analysis. Previously, we reported a NTRK fusion prevalence of 22% in this subgroup [18]. Transcription of NTRK fusions detected by FFPE-TLC was confirmed by RNA-NGS.

RNA-NGS

The Archer Fusionplex® Lung panel (ArcherDX, San Diego, USA) was used to assess NTRK1, NTRK2, and NTRK3 rearrangements (Supplementary Table S1). RNA was isolated using the Maxwell® RSC 48 instrument and the Maxwell® RSC RNA FFPE kit (Promega, Madison, USA) from four FFPE sections of 10 µm with a minimum of 10% tumor cell content. Isolated RNA was quantified using the Quantus™ Fluorometer and the QuantiFluor®RNA system (Promega, Madison, USA). A quantitative PCR (qPCR)-based method was used to determine the quality of the mRNA in each sample prior to the targeted library preparation. When a sample did not meet the quality criteria (Cq quantitative PCR should be < 30 cycles), the isolation and quality control was repeated using slides from a different tumor tissue block (if available). Libraries were prepared using the Archer FusionPlex reagent kit for Illumina (ArcherDX, San Diego, USA) according to manufacturer’s instructions. Results were analyzed on the ArcherDX website (https://analysis.archerdx.com/), version 6.2 or 7.0. Criteria for analysis were minimum of 200.00 RNA reads, unique RNA start sites per GSP2 control ≥ 10, on target deduplication ratio ≥ 3 and ≤ 10. Fusions must have > 3 SS unique start sites, > 5 breakpoint spanning reads, 10% fusion reads, and must be in frame. All results were examined by the same molecular biologist (WWL), who was blinded for the IHC result.

FFPE-targeted locus capture

A detailed description of FFPE-TLC was published before [15]. In brief, two to four FFPE sections of 10-µm thickness were deparaffinized to enable in situ DNA digestion by a restriction enzyme. After in situ ligation and overnight reverse crosslinking, standard protocols for library preparation and hybridization capture-based target enrichment were followed. Resulting libraries were paired-end sequenced on an Illumina platform. Two different capture panels were used to target NTRK1, NTRK2, NTRK3, and several other genes, including BRAF. For a subset of samples KRAS, NRAS, EGFR, and PIK3CA were also covered (Supplementary Table S2). One million on-target reads were aimed for 1 Mb region of interest.

After alignment of the raw sequencing data to the human genome, a computational pipeline PLIER was run to automatically detect NTRK1, NTRK2, and NTRK3 gene fusions. In brief, PLIER detects genomic intervals with significantly increased coverage of proximity ligation products per target locus and calculates a z-score by comparing its observed proximity score to the related expected proximity score. This PLIER pipeline has been optimized based on lymphoma cases that nearly always present as interchromosomal (trans) fusions, with fusion partners on a different chromosome. For CRC, several intrachromosomal (cis) fusions have been described, like for example LMNA::NTRK1, with both genes being located on chromosome 1. Therefore, visual inspection was performed in addition to PLIER to detect these intrachromosomal rearrangements. The scientists (ES, JFS, HF) involved in rearrangement calling were blinded for the results of the other assays.

Fluorescence in situ hybridization

FISH was performed on all pan-TRK positive cases [19]. NTRK1, NTRK2, and NTRK3 break-apart FISH probes (Z-2167, Z-2205, and Z-2206, ZytoVision GmbH, Bremerhaven, Germany) were used in a stepwise approach. First, the NTRK1 break-apart probe was used, as the NTRK1 gene is the most common NTRK fusion detected in colorectal cancer. Negative cases were subsequently analyzed using the NTRK3 break-apart probe. Lastly, FISH using the NTRK2 probe was performed for cases that were negative for both the NTRK1 and NTRK3 FISH.

For all tumor samples, a minimum of 100 non-overlapping nuclei were scored for their signals. FISH was considered positive for an NTRK fusion if ≥ 10% tumor cells showed a separation of red and green signals with a minimum of two signal diameters. With respect to single red signals: FISH was considered positive if ≥ 15% tumor cells showed a single red signal and equivocal if 10–15% of tumor cells showed a single red signal. FISH analysis was performed by qualified analysts and interpreted by a molecular biologist (WWL).

5′/3′ imbalance quantitative RT-PCR (qRT-PCR)

For the qRT-PCR we have used the Idylla GeneFusion Assay (Biocartis NV, Mechelen, Belgium). One 10-µm-thick slide was used according to the manufacturer’s instructions. A result was obtained for NTRK1, NTRK2, and NTRK3 individually. In case of errors or invalid test results, the assay was repeated once with another slide from the same tissue block. An invalid result for one out of three NTRK genes in the absence of a detected NTRK fusion was scored as invalid.

Comparison between assays

For all assays, the robustness was determined. Robustness was defined as the percentage of analyzed samples that provided an interpretable result in the first run. The assay results were all reviewed by two independent molecular biologists (WWL and RJAF), and their consensus decision regarding the presence of NTRK fusions per sample was used as truth set reference. Using this truth set reference, sensitivity and specificity were calculated for RNA-NGS, FFPE-TLC, FISH, and qRT-PCR. Subsequently, detected mutations in BRAFV600E and KRAS by RNA-NGS and FFPE-TLC were compared, and the level of agreement was evaluated.

Statistical analysis

Statistical analysis was performed using R Statistical Software (v4.0.1) and the epiR package. P-values of < 0.05 were considered statistically significant.

留言 (0)

沒有登入
gif