CINSARC signature outperforms gold-standard TNM staging and consensus molecular subtypes for clinical outcome in stage II-III colorectal carcinoma.

The outcome of stage II-III colorectal cancer (CRC) is highly variable and therapeutic choice is currently based on TNM staging with a few additional biomarkers. However, studies show that some stage III patients have a better prognosis than some stage II patients. A promising consensus molecular (CMS) classification with prognostic relevance has been developed, but it is not used in daily practice. Our team developed CINSARC, a 67-gene expression prognostic signature, whose prognostic value has been demonstrated in many cancer types. It is applicable to formalin-fixed, paraffin-embedded (FFPE) blocks using NanoString® technology. We investigated whether it could predict outcome in stage II-III CRC. We established the CINSARC classification on the TCGA retrospective cohort comprising 297 stage II-III CRC patients using RNA sequencing and on a second independent cohort comprising 169 cases using NanoString® technology. We compared its recurrence-free and overall survival prognostic value with TNM staging and CMS classification. In the TCGA cohort, we showed that CINSARC significantly splits the population of stage II-III CRC into two groups with different progression-free interval (P = 1.68 × 10-2; HR = 1.87 [1.11-3.16]) and overall survival (P = 3.73 × 10-3; HR = 2.45 [1.31-4.59]) and is a strong prognostic factor in multivariate analysis, outperforming TNM staging and CMS classification. We validated these results in the second cohort by applying CINSARC on FFPE samples with Nanostring® technology. CINSARC is a ready-to-use tool with a robust independent prognostic value in stage II-III CRC.

An unusual phenotype occurs in 15% of mismatch repair-deficient tumors and is associated with non-colorectal cancers and genetic syndromes.

Immunohistochemistry (IHC) and/or MSI-PCR (microsatellite instability-polymerase chain reaction) tests are performed routinely to detect mismatch repair deficiency (MMR-D). Classical MMR-D tumors present a loss of MLH1/PMS2 or MSH2/MSH6 with MSI-High. Other profiles of MMR-D tumors have been described but have been rarely studied. In this study, we established a classification of unusual MMR-D tumors and determined their frequency and clinical impact. All MMR-D tumors identified between 2007 and 2017 were selected. Any profile besides the classical MMR-D phenotype was defined as unusual. For patients with unusual MMR-D tumors, IHC, and PCR data were reviewed, the tumor mutation burden (TMB) was evaluated and clinical and genetic features were collected. Of the 4948 cases of MMR testing, 3800 had both the available IHC and MSI-PCR results and 585 of these had MMR-D. After reviewing the IHC and PCR, 21% of the cases initially identified as unusual MMR-D were reclassified, which resulted in a final identification of 89 unusual MMR-D tumors (15%). Unusual MMR-D tumors were more often associated with non-CRC than classical MMR-D tumors. Unusual MMR-D tumors were classified into four sub-groups: i) isolated loss of PMS2 or MSH6, ii) classical loss of MLH1/PMS2 or MSH2/MSH6 without MSI, iii) four MMR proteins retained with MSI and, iv) complex loss of MMR proteins, with clinical characteristics for each sub-group. TMB-high or -intermediate was shown in 96% of the cancers studied (24/25), which confirmed MMR deficiency. Genetic syndromes were identified in 44.9% (40/89) and 21.4% (106/496) of patients with unusual and classical MMR-D tumors, respectively (P < 0.001). Five patients treated with an immune checkpoint inhibitor (ICI) had a prolonged clinical benefit. Our classification of unusual MMR-D phenotype helps to identify MMR deficiency. Unusual MMR-D phenotype occurs in 15% of MMR-D tumors. A high frequency of genetic syndromes was noted in these patients who could benefit from ICI.

High-throughput detection of clinically targetable alterations using next-generation sequencing.

Next-generation sequencing (NGS) has revolutionized the therapeutic care of patients by allowing high-throughput and parallel sequencing of large numbers of genes in a single run. However, most of available commercialized cancer panels target a large number of mutations that do not have direct therapeutic implications and that are not fully adapted to low quality formalin-fixed, paraffin-embedded (FFPE) samples. Here, we designed an amplicon-based NGS panel assay of 16 currently actionable genes according to the most recent recommendations of the French National Cancer Institute (NCI). We developed a panel of short amplicons (<150 bp) using dual-strand library preparation. The clinical validation of this panel was performed on well-characterized controls and 140 routine diagnostic samples, including highly degraded and cross-linked genomic DNA extracted from FFPE tumor samples. All mutations were detected with elevated inter-laboratory and inter-run reproducibility. Importantly, we could detect clinically actionable alterations in FFPE samples with variant allele frequencies as low as 1%. In addition, the overall molecular diagnosis rate was increased from 40.7% with conventional techniques to 59.2% with our NGS panel, including 41 novel actionable alterations normally not explored by conventional techniques. Taken together, we believe that this new actionable target panel represents a relevant, highly scalable and robust tool that is easy to implement and is fully adapted to daily clinical practice in hospital and academic laboratories.