Introduction: Colorectal carcinoma (CRC) is among the most common carcinomas in women and men. In the advanced stage, patients are treated based on the RAS status. Recent studies indicate that in the future, in addition to KRAS and NRAS, alterations in other genes, such as PIK3CA or TP53, will be considered for therapy. Therefore, it is important to know the mutational landscape of routinely diagnosed CRC. Method: We report the molecular profile of 512 Swiss CRC patients analyzed by targeted next-generation sequencing as part of routine diagnostics at our institute. Results: Pathogenic and likely pathogenic variants were found in 462 (90%) CRC patients. Variants were detected in TP53 (54.3%), KRAS (48.2%), PIK3CA (15.6%), BRAF (13.5%), SMAD4 (10.5%), FBXW7 (7.8%), NRAS (3.5%), PTEN (2.7%), ERBB2 (1.6%), AKT1 (1.5%), and CTNNB1 (0.9%). The remaining pathogenic alterations were found in the genes ATM(n= 1), MAP2K1(n= 1), and IDH2(n= 1). Discussion/Conclusions: Our analysis revealed the prevalence of potential predictive markers in a large cohort of CRC patients obtained during routine diagnostic analysis. Furthermore, our study is the first of this size to uncover the molecular landscape of CRC in Switzerland.

© 2022 The Author(s). Published by S. Karger AG, Basel

Introduction

Colorectal cancer (CRC) is one of the most common cancers worldwide and the second most prevalent cancer in Europe [1]. CRC death rates have been steadily decreasing for the last decades, mainly due to increased screening with the removal of precancerous colorectal polyps, improved diet, and the advent of targeted therapy [1].

Approximately 50% of CRC patients have stage I or II disease at initial diagnosis and are usually treated only surgically [1]. In contrast, patients with metastatic disease (stage III or IV) receive adjuvant therapy, which includes a combination of cytotoxic and biological targeted agents with or without ablative techniques [2, 3]. Targeted therapy with anti-epidermal growth factor receptor (EGFR) monoclonal antibodies like cetuximab and panitumumab has greatly improved clinical outcomes of metastatic CRC [4]. Yet, patients with KRAS or NRAS variants will not profit from anti-EGFR therapy [2]. Moreover, recent data indicate that patients with BRAF V600E mutations will not profit from either cetuximab or panitumumab as single agent therapy or in combination with cytotoxic chemotherapy [3]. In addition, patients with BRAF V600E mutations have a poor prognosis. As a result, the European and American cancer associations recommend testing advanced CRC patients for variants of KRAS and NRAS in exon 2, 3, and 4 and BRAF V600E [2, 3].

Mutational testing is usually performed by next-generation sequencing (NGS), which has revolutionized clinical practice and allows screening of these mutations with high accuracy [5, 6]. Moreover, due to its convenience, many genes besides KRAS, NRAS, and BRAF are commonly analyzed by NGS. Recent data seem to indicate that other genetic changes may be associated with a lack of response to anti-EGFR therapy, such as PIK3CA mutations, phosphatase and tensin homolog (PTEN) loss, and HER2 amplification [7-9]. However, previous evidence from randomized trials has not generated strong evidence that these genes should be analyzed before anti-EGFR therapy [10]. Despite their current lack of clinical significance, it is expected that genes such as PIK3CA, HER2, PTEN, but also AKT1, TP53, and MAP2K1 may play a role in future therapeutics, due to newly developed targeted therapy [11].

Therefore, it is important to know which variants can be detected besides KRAS, NRAS, and BRAF in routine CRC patient samples. Consequently, we have performed a retrospective study and summarized the molecular alterations detected in a cohort of CRC (n = 512) samples, analyzed by NGS at our Institute of Pathology within a 4-year period. Besides the discovery of rare tumor-specific variants, our study also allowed us to obtain the mutational landscape of CRC samples in Switzerland.

Materials and MethodsSamples

We searched for tissue samples from patients with CRC which underwent mutational analysis by NGS between January 2015 and January 2019 at the Institute of Pathology of the University Hospital Basel. All tissue samples were formalin fixed and paraffin (FFPE) embedded. Eighty-two percent (n = 420) were fine needle biopsies obtained during endoscopy and 18% (n = 92) were received from surgical resections. NGS analysis was performed in 82% (n = 420) on the primary tumor sample and in 18% (n = 92) on metastatic samples. The study was approved by the Ethics Commission of Northern Switzerland (EKNZ; study ID: 2019-00776).

Macroscopic and Microscopic Analysis

Each tumor was evaluated by a board-certified pathologist, who evaluated the differentiation grade (low grade or high grade) and the morphology of the primary tumor (conventional type, mucinous, medullary, or signet-ring cell type) according to the 2019 WHO classification of tumors of the digestive system [12]. Expression of mismatch repair proteins (MLH1, PMS2, MSH2, MSH6) was analyzed by immunohistochemistry. The following antibodies were used for the analysis: MLH1 from Ventana/Roche (790-4535) or Leica biosystem (ES05), MSH2 from Ventana/Roche (760-5093) or Cell marque (G219-1129), MSH6 from Ventana/Roche (760-5092) or Cell marque (SP93), PMS2 from Dako (IR087) or Cell marque (EPR 3947).

Tissue Selection and DNA Extraction

For DNA isolation, tissue blocks were cut to sections of 4 μm thickness, which were placed on a glass slide. Afterward, the area of interest was marked by a pathologist using an H&E-stained tissue section as a guide. The selected area was macrodissected, and DNA was isolated using the Maxwell RSC FFPE Plus DNA kit (Catalog number: AS1720) according to the manufacturer’s protocol. DNA was quantified by the Qubit dsDNA HS kit (Thermo Fisher Scientific, Cat.No. Q32854).

Next-Generation Sequencing

423 sample (83%) were analyzed by the Ion TorrentTM Oncomine SolidTM Tumor Panel covering the following 22 genes with their exons (in brackets): AKT1 (3), ALK (22, 23, 25), BRAF (11, 15), CTNNB1 (3), DDR2 (5, 8, 12, 13, 14, 15, 17), EGFR (12, 18, 19, 20, 21), ERBB2 (19, 20, 21), ERBB4 (3, 4, 6, 7, 8, 9, 15, 23), FBXW7 (5, 8, 9, 10, 11), FGFR1 (4, 7), FGFR2 (7, 9, 12), FGFR3 (7, 9, 14, 16, 18), KRAS (2, 3, 4), MAP2K1 (2), MET (2, 14, 16, 19), NOTCH1 (26, 27), NRAS (2, 3, 4), PIK3CA (10, 14, 21), PTEN (1, 3, 6, 7, 8), SMAD4 (3, 5, 6, 8, 9, 10, 11, 12), STK11 (1, 4, 5, 6), TP53 (2, 4, 5, 6, 7, 8, 10). Eighty-nine samples (17%) were analyzed using the Ion AmpliSeqTM Cancer Hotspot Panel v2 covering the following 50 genes and their hotspots: ABL1, AKT1, ALK, APC, ATM, BRAF, CDH1, CDKN2A, CSF1R, CTNNB1, EGFR, ERBB2, ERBB4, EZH2, FBXW7, FGFR1, FGFR2, FGFR3, FLT3, GNA11, GNAS, GNAQ, HNF1A, HRAS, IDH1, IDH2, JAK2, JAK3, KDR, KIT, KRAS, MET, MLH1, MPL, NOTCH1, NPM1, NRAS, PDGFRA, PIK3CA, PTEN, PTPN11, RB1, RET, SMAD4, SMARCB1, SMO, SRC, STK11, TP53, VHL.

Library Preparation

The Ion AmpliSeq Library Kit 2.0 or Plus (Thermo Fisher Scientific) was used to prepare the libraries from 10 ng of DNA using the Oncomine Solid Tumor Panel or the Ampliseq Cancer Hotspot Panel v2. First, the Ion Ampliseq HiFi Master Mix (Thermo Fisher Scientific) was used to prepare the amplicons that were digested with FUPA reagent and tagged with barcode adapters. Next, the amplified products were purified using the Agencourt AMPure XP PCR purification system (Beckman Coulter, California, USA). The purified libraries were then diluted 1:1,000 and quantified by qPCR using the Ion Universal Quantitation Kit (Thermo Fisher Scientific). The quantified stock libraries were then diluted to 50pM for downstream template preparation on the Ion Chef instrument (Thermo Fisher Scientific). NGS libraries were sequenced on an Ion S5TM instrument (Thermo Fisher Scientific) using semiconductor sequencing technology. Sequencing runs were planned on the Torrent Suite SoftwareTM v5.6 or later, and barcoded libraries were pooled and loaded on an Ion 530TM or 540TM chip using the Ion ChefTM instrument (Thermo Fisher Scientific). The loaded chip was then inserted into the initialized Ion S5XLTM instrument and sequenced using 500 flows. Raw data were processed automatically on the Torrent ServerTM and aligned to the reference hg19 genome. QC was performed manually for each sample, aiming for the following metrics; on-target reads >90%, read uniformity >90%, and mean read depth >2,000. The sequencing data of the QC passing samples were then uploaded in BAM format to the Ion ReporterTM Analysis Server for variant calling and annotation.

Data Analysis and Statistical Analysis

Variant detection was performed on the Ion ReporterTM Analysis Software v5.6 or later (Thermo Fisher Scientific) using the appropriate workflow for either the Oncomine Solid Tumor Panel or Ampliseq Cancer Hotspot Panel v2. After variant calling, a filter chain was applied to remove polymorphisms based on cross-referencing with UCSC common SNPs, ExAC, 10,000 Genomes, and 5,000 Exomes databases. Furthermore, remaining variants were filtered based on Phred quality score >100, allele read depth >350, strand bias <90%, minimal allelic frequency >5%. Lastly, sequence variants were evaluated for their pathogenicity based on previous literature, databases (COSMIC, ClinVar, OncoKB, Varsome), and by using the open-access version from the Cancer Core Europe online portal [13]. Mutations were classified as pathogenic, likely pathogenic, variant of unknown significance (VUS), likely benign, and benign. Mutations classified as benign or likely benign were not reported. For statistical analysis, Fisher’s exact test was used. Statistical analyses were performed using R software package version 3.6.0 (www.r-project.org).

ResultsPatients and Morphological Analysis

A total of 512 CRC patients, 208 females (41%) and 304 (59%) males, were included (shown in Table 1). The average age at diagnosis was 67 years (standard deviation ±12). The tumor was located in the left colon (descending colon, sigmoid colon, and rectum) in 59% (n = 300) and in the right colon (cecum, ascending colon, and transverse colon) in 34% (n = 175). In 7% (n = 37), the primary tumor localization remained unknown. Tumor tissue analyzed by NGS was obtained from 420 (82%) primary tumors and 92 (18%) metastases. Among the metastases, 61 (67%) were from the liver, 17 (19%) from the lungs, 6 (7%) from the peritoneum, 5 (5%) from the lymph nodes, and 3 (3%) from the bone. Most patients (96%; n = 492) were diagnosed with stage III-IV disease, and only few patients with stage I-II (4%; n = 20). Forty-one of all patients (8%) received neoadjuvant chemotherapy before tissue collection, and all these samples were from patients with metastatic diseases. Morphological analysis of all primary tumors showed in 78% (n = 326) low-grade differentiation and in 22% (n = 94) high-grade differentiation, whereas metastatic tumors showed in 75% (n = 69) low-grade differentiation and in 25% (n = 23) high-grade differentiation. Most CRC showed conventional morphology (95.5%; n = 489), whereas 18 (3.5%) showed mucinous morphology, 4 (0.8%) signet-ring cell morphology, and 1 (0.2%) medullary morphology.

Table 1.

Clinical characteristics of patient cohort (n = 512)

Variant Analysis

Data from CRC patients sequenced by NGS over a 4-year period were analyzed. At the beginning of the 4-year period, CRC samples were sequenced using the Ion AmpliSeqTM Cancer Hotspot Panel v2 (17%). Afterward, for convenience, this panel was replaced by the Ion TorrentTM Oncomine SolidTM Tumor Panel. In summary, 462 (90%) CRC showed one or more pathogenic or likely pathogenic variant, and 5 (1%) CRC only one VUS. A total of 45 (9%) CRC did not show a pathogenic or likely pathogenic variant, or a VUS. Pathogenic or likely pathogenic variants were detected in TP53 (54.3%), KRAS (48.2%), PIK3CA (15.6%), BRAF (13.5%), SMAD4 (10.5%), FBXW7 (7.8%), NRAS (3.5%), PTEN (2.7%), ERBB2 (1.6%), AKT1 (1.5%), and CTNNB1 (0.9%). The remaining pathogenic alterations were found in the genes ATM (n= 1), MAP2K1(n= 1), and IDH2(n= 1).

Around a third of the CRC (38%, n = 193) showed only one variant among the genes sequenced. Variants in two or three genes were found in 175 (34%) and 73 CRC (14%), respectively. Twenty-one CRC (4%) showed four or more variants in the genes analyzed. The majority of the alterations were missense (88.0%) or nonsense (8.1%) variants, whereas the remaining alterations consisted in frameshifts (2.8%), inframe indels (1.0%), and splice sites (0.2%) (shown in Fig. 1; Table 2; online suppl. Table 1 for a complete list of the detected variants; see www.karger.com/doi/10.1159/000526117 for all online suppl. material).

Table 2.

Pathogenic and likely pathogenic variants in potentially actionable genes

Fig. 1.

Overview of clinical and molecular features. Each column represents a patient with clinical features annotated on top of the oncoPrint (gender, tumor origin, tumor side, MS status). Mutational frequency is shown per gene. Molecular alterations are separated in missense, nonsense, frameshifts, inframe indels, and splice site variants. Columns are sorted by gender, origin, side, and MS status.

More than half of the CRC had variants affecting the EGFR/RAS/RAF/MAPK pathway. The most frequent variants were in KRAS (48.2%; n = 247) followed by BRAF (13.5%; n = 69). KRAS variants were located in 219 (42.8%) CRC on exon 2, in 14 (2.7%) on exon 3, and in 14 (2.7%) on exon 4. The most common KRAS variant was p.G12D (11.9%; n = 61), followed by p.G13D (10.9%; n = 56), and p.G12V (9.8%, n = 50). The KRAS p.G12C variant was present in only 16 CRC (3.1%). One CRC had a double KRAS variant located on codon p.G12V and p.A146V. Six CRC had concomitant variants on KRAS and BRAF. However, none of these samples showed the BRAF variant p.V600E. Also, the allelic frequency of the KRAS and BRAF variants was similar, possibly suggesting that these mutations occurred within the same clone. NRAS variants were found in 18 CRC (3.5%) and were located in 10 CRC (2%) on exon 3 and in 8 (1.6%) on exon 2. A total of five CRC had concomitant variants in NRAS and BRAF at similar allelic frequency. Of them, only one showed a combination with the BRAF p.V600E variant.

BRAF variants were located mostly on exon 15 (12.1%; n = 62) and on exon 11 (1.4%; n = 7). Most variants were located in the p.V600E hotspot region (10.0%; n = 51), while 18 CRC (3.5%) showed variants outside of this specific hotspot. Of all BRAF variants, 52 fall into class I, 3 into class II, and 14 into class III [14]. Interestingly, as described above, although the BRAF p.V600E mutation was the most frequently observed, it only co-occurred in one CRC with a NRAS variant and never co-occurred with a KRAS variant.

Targetable variants in the phosphatidylinositol 3-kinase (PI3K) pathway genes (PIK3CA, AKT, PTEN) were observed in 102 (19.9%) of all CRC. PIK3CA variants were found in 15.6% (n = 80) CRC. The most frequent PIK3CA variant was p.E545K (5.5%; n = 28) followed by p.E542K (2.1%; n = 11). PIK3CA gene variants were associated in 94% (n = 78) with concomitant variants in other driver genes such as KRAS (65%, n = 51) or BRAF (17%, n = 13). PIK3CA gene mutations alone were found only in five CRC. AKT1 variants were found in a total of 8 CRC (1.6%), and the most frequent variant within all AKT variants was p.E17K (n = 6). All except for one CRC had concomitant variants in other driver genes besides the AKT1 variant. PTEN was mutated in 14 (2.7%) CRC and occurred in combination with other mutated genes in all but one CRC.

The APC gene, a part of the WNT-beta-catenin pathway, was only examined for hotspot mutations in tumors sequenced with the Ion AmpliSeqTM Cancer Hotspot Panel v2 (n = 89). Variants were observed in 31.4% (n = 28) of the samples analyzed with this specific panel. CTNNB1 variants were detected in 1.0% (n = 5) and were always concomitant with other variants, mostly KRAS (n = 2).

In addition, we found that variants involving tumor suppressor genes were detected mainly in TP53 (54.3%; n = 278) and FBXW7 (7.8%; n = 40). One CRC was found with an ATM variant. Inactivating variants in the TGF-β pathway were observed in the gene SMAD4 (10.5%; n = 54). Molecular alterations involving specific receptor tyrosine kinases (RTK) such as ERBB2 (1.6%; n = 8) were also found. Finally, one CRC had a pathogenic variant in IDH2 and one in MAP2K1.

Microsatellite Status

Microsatellite status was tested in 410 (80%) tumors. 99% (n = 406) were tested by immunohistochemistry for the DNA mismatch repair proteins (MLH1, PMS2, MSH2, and MSH6) and 1% (n = 4) by PCR using the Bethesda panel. 94% (n = 385) were microsatellite stable (MSS) and 6% (n = 25) were microsatellite instable (MSI). MSI CRC most commonly showed an immunohistochemical loss of MLH1 and PMS2 (23 CRC; 93%), whereas one demonstrated a solitary loss for MSH6 and one showed a loss of MSH2 with MSH6. Molecular analysis of the MSI CRC revealed BRAF variants in 61% (n = 14).

Gene Alterations Associated with Clinical Parameters

Finally, we performed statistical analysis to determine correlations of variant status with clinical parameters. We found that TP53 and KRAS mutations are associated with MSS tumor status (Fisher’s exact-adjusted p value = 0.0048 and 0.0322), whereas BRAF and FBXW7 gene mutations are most commonly observed in correlation with MSI tumor status (Fisher’s exact-adjusted p value = 0.0009 and 0.0175) (shown in Fig. 2, left panel). In relation to the location of the tumor, BRAF was associated with right-side location (Fisher’s exact-adjusted p value = 4.4e−10) (shown in Fig. 2, right panel).

Fig. 2.

Gene alterations associated with clinical features. Plots show log2 odds ratios (OR, x-axis) against −log2-adjusted p values (y-axis) obtained using Fisher’s exact test (adjusted for multiple hypothesis testing using the FDR method). Mutational frequency of specific genes is significantly associated with MSI status (left panel) or tumor localization (right panel), as shown when above the dotted line (−log2 (p value) = 4.33, adjusted p value = 0.05).

In addition, we wanted to know if the total number of mutations detected was different between primary and metastatic CRC. For comparison, we separated tumors based on the microsatellite stability status. Interestingly, we did not detect a difference in the total number of mutations between primary and metastatic CRC (shown in Fig. 3a). Similarly, we compared the number of mutations between neoadjuvantly treated and untreated CRC. We selected only MSS CRC because the number of samples was too small for MSI CRC. Again, we did not detect a difference in the number of mutations between these two groups (shown in Fig. 3b).

Fig. 3.

Comparison of number of detected mutations between primary and metastatic CRC as well as treated and untreated CRC. a Plots show comparison between primary and metastatic CRC after grouping according to microsatellite status. b Plots show comparison between chemotherapeutically treated and untreated (N/A) CRC. Fisher’s exact test was used for comparison.

Discussion

To our best knowledge, this is the first study which analyzes the molecular profile of CRC in a Swiss cohort. We found pathogenic or likely pathogenic variants in 90% of CRC. The highest prevalence was detected in TP53, followed by KRAS, APC, PIK3CA, and BRAF. Similar studies have been performed in the USA and other European countries, where a comparable mutational landscape in their population was found [15-18]. Indeed, TP53 is usually one of the most frequently affected genes in CRC with a prevalence similar to ours of 54.3% [15-17]. Variants in the RAS/RAF/MAPK pathway are also frequently found in CRC, and earlier studies in Western countries have shown that KRAS variants were present between 37.8%, 46.1%, and 49.3% [16-18]. In line with these datasets, our cohort revealed KRAS variants in 48.2%, with the majority of mutations located in exon 2, and less frequently in exon 3 and 4. The KRAS p.G12C variant, predictive for response to KRAS p.G12C inhibitors [19], was only present in 16 CRC, corresponding to 3.1% of the whole cohort. BRAF variants were found in 13.5% CRC of our cohort, which is slightly higher to previous studies, where BRAF variants were found in 10.5%, 6.5%, and 9.6% [16-18]. These differences might be explained by differences in tumor localization and microsatellite stability status. As expected, BRAF p.V600E was the most frequent variant detected. Despite the previously hypothesized exclusivity of RAS and BRAF variants [20], we found six CRC with concomitant variants on KRAS and BRAF, and five with concomitant variants on NRAS and BRAF. Among all concomitant BRAF variants, we found one with a NRAS and co-occurring BRAF p.V600E variant. Earlier studies have found co-occurrence with BRAF p.V600E, and they also reported a frequent combination of RAS variants with non-V600E BRAF variants [21]. Among RAS family genes, mutations in NRAS were with a prevalence of 3.5% less frequent compared to KRAS variants, confirming data from previous studies [22, 23].

Variants in the PI3K/AKT pathway are potentially actionable. PIK3CA, the catalytic subunit of PI3K, has been found mutated in CRC previously at 14%, 15%, and 20.3% [17, 18, 24]. Accordingly, our cohort showed a 15.6% rate for PIK3CA variants. One important component of the PI3K/AKT pathway is PTEN, a tumor suppressor gene, which we found to be mutated in 2.7% CRC.

Activation of the WNT-beta-catenin pathway by CTNNB1 variants was detected in 0.9% of our cohort, confirming data from previous studies [18, 25]. In all CRC, CTNNB1 variants were associated with at least one other concurrent variant, most commonly in KRAS, potentially indicating that CTNNB1 variants are a late event in CRC carcinogenesis [18]. The prevalence of APC variants was 31.4%. However, only a portion of our cohort was profiled for hotspot mutations in APC (only samples analyzed with the Ion AmpliSeqTM Cancer Hotspot Panel v2). Therefore, no conclusive statement can be made for APC mutations within our cohort. Indeed, previous studies described APC variants to be mutated in up to 80% [15]. In our study, the prevalence of SMAD4-mutated CRC was 10.5%, which is in line with previous studies showing that SMAD4 mutations occur in 2–20% of CRC [18, 26]. Because SMAD4 mutation is associated with clinicopathological parameters such as tumor localization, disease stage, or RAS status, the selection of the cohort may influence the prevalence of SMAD4-mutated CRC. Presence of SMAD4 mutations may be further associated with a worse prognosis and resistance to 5-fluorouracil chemotherapy, probably through activation of the AKT pathway [26, 27]. FBXW7 is a tumor suppressor gene and was mutated in 7.8% of our cohort, similar to previous studies [28]. Recent data suggest that inactivating variants of FBXW7 could predict response to mTOR inhibitor rapamycin [29]. However, additional molecular alterations, for example, in KRAS, may contribute to the limited therapeutic efficacy of mTOR inhibitors [18]. Indeed, in our cohort, 45.2% of the FBXW7-mutated CRC had concomitant variants in the KRAS gene. Our cohort further revealed that ERBB2 was mutated in 1.6% of the CRC, which is a little bit higher than in earlier studies [16, 18]. Because recent data have suggested that ERBB2 variants may constitute a mechanism of resistance to EGFR antibodies, this observation may become relevant for therapeutic management [30]. Finally, we found one CRC with an IDH2 variant and one with a MAP2K1. Both mutations are targetable in other tumors; however, data on the effectiveness of these drugs on CRC patients are currently scarce [31, 32].

Statistical analysis of our data showed that TP53 and KRAS mutations variants were more frequently associated with MSS tumor status. Moreover, BRAF variants were associated with right tumor localization and MSI status. These correlations have also been shown in previous studies [33, 34]. We did not observe a difference in the number of detected variants between primary tumor samples and metastases. The interpretation of our data has to be made with caution because we only performed targeted sequencing and did not analyze matched primary and metastatic samples. Nevertheless, it confirms earlier data, which suggested that most mutations are present similarly between primary and metastatic samples, both in terms of number and type of mutations [35, 36]. Similarly, we did not detect a difference between neoadjuvant-treated CRC samples and untreated samples. Although limited data are available, it seems that neoadjuvant therapy does not impact the mutational signature of CRC [37].

In conclusion, our study characterized the distribution of targetable molecular alterations in a large cohort of Swiss CRC patients and provides meaningful information in a clinical setting. Our results are in line with other studies, indicating that geographical factors do not play a substantial role in mutagenesis for CRC.

Acknowledgments

We would like to thank Anja Förster, Sibylle Tschumi, Tanja Dietsche, Valeria Perrina, and Linda Schweizer for excellent technical assistance. We also would like to thank Sarah Niederberger for her outstanding support and Jasmin Haslbauer for critically reading the manuscript.

Statement of Ethics

The study was approved by the Ethics Commission of Northern Switzerland (EKNZ; study ID: 2019-00776). Because this study included only data from samples obtained as part of routine clinical diagnosis, written informed consent from patients was not required in accordance with local/national guidelines. However, patients were excluded if they declined research on their tissue or on their clinical data. The research was conducted ethically in accordance with the World Medical Association Declaration of Helsinki.

Conflict of Interest Statement

Matthias S. Matter has served as a consultant for Novartis and Glaxo Smith Cline and received speaker’s honoraria from Thermo Fisher and Merck, all outside the current work. The other authors, namely, Simon Haefliger, Katharina Marston, Ilaria Alborelli, Philip M. Jermann, Edouard-Jean Stauffer, Mathias Gugger, Sylvia Hoeller, Luigi Tornillo, Luigi M. Terracciano, and Michel Bihl have no conflicts of interest to declare.

Funding Sources

Swiss Cancer Research Foundation Grant KFS-4168-02-2017 and Swiss National Science Foundation (SNSF; Grant No. 320030_189275) to Matthias S. Matter. The sponsor of the study did not have any role in the study design or collection, analysis, and interpretation of data.

Author Contributions

Simon Haefliger, Katharina Marston, and Matthias S. Matter conceived and designed the project and wrote the manuscript. Michel Bihl analyzed the NGS data and reported the mutations. Simon Haefliger, Katharina Marston, Ilaria Alborelli, Philip M. Jermann, Edouard-Jean Stauffer, Mathias Gugger, Sylvia Hoeller, Luigi Tornillo, Luigi M. Terracciano, and Matthias S. Matter performed histological analysis and interpreted the sequencing data. Ilaria Alborelli performed molecular data visualization. All authors revised and agreed to the content of the manuscript.

Data Availability Statement

All data generated or analyzed during this study are included in this article and its online supplementary material. Further inquiries can be directed to the corresponding author.

This article is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC). Usage and distribution for commercial purposes requires written permission. Drug Dosage: The authors and the publisher have exerted every effort to ensure that drug selection and dosage set forth in this text are in accord with current recommendations and practice at the time of publication. However, in view of ongoing research, changes in government regulations, and the constant flow of information relating to drug therapy and drug reactions, the reader is urged to check the package insert for each drug for any changes in indications and dosage and for added warnings and precautions. This is particularly important when the recommended agent is a new and/or infrequently employed drug. Disclaimer: The statements, opinions and data contained in this publication are solely those of the individual authors and contributors and not of the publishers and the editor(s). The appearance of advertisements or/and product references in the publication is not a warranty, endorsement, or approval of the products or services advertised or of their effectiveness, quality or safety. The publisher and the editor(s) disclaim responsibility for any injury to persons or property resulting from any ideas, methods, instructions or products referred to in the content or advertisements.