This article is mentioned in:

Introduction

Colorectal cancer (CRC) is the second leading cause of cancer death worldwide (1). The estimated number of deaths due to CRC in Europe was 243,000 in 2018(2). Approximately 25% of patients present metastases at the time of diagnosis (synchronous disease) and about half of the remaining patients will develop metachronous metastases, contributing to the high mortality rates reported for CRC (3).

The treatment of metastatic CRC (mCRC) is based on chemotherapy when metastases cannot be removed by surgery. Treatment is guided by molecular information. Microsatellite unstable tumors are treated with immunotherapy (4), while microsatellite stable tumors, which account for 95% of mCRC, are treated by cytotoxic agents and targeted therapies. The principal cytotoxic agents currently used are fluoropyrimidines, irinotecan and oxaliplatin. The chemotherapy regimen often consists of fluorouracil and folinic acid combined with either oxaliplatin (FOLFOX) or irinotecan (FOLFIRI) or both (FOLFOXIRI). These chemotherapies are associated with two different classes of target therapies: Anti-angiogenesis drugs, and anti-epidermal growth factor receptor (EGFR). While antiangiogenics can be used broadly (5,6), anti EGFR therapy is only effective in patients with RAS wild type status (3,7).

It is currently established that mCRC is a heterogeneous disease. Tumor sidedness carries strong prognostic value (8) while synchronous mCRC is associated with a particularly poor prognosis. These data suggest that the biological characteristics of these tumors are different. Several papers have introduced CRC molecular subtyping systems, which are currently summarized in consensus molecular subtypes (9). This transcriptomic typology classifies patients into 4 consensual subtypes with different prognostic behavior. Using large panel sequencing data, a recent analysis detailed the molecular landscape of mCRC (10). This study observed different gene variations between left and right-sided tumors. However, the genomic structural pattern was not assessed in this study.

Here, using exome analysis of 77 mCRC patients, we aimed to determine genomic structural patterns and gene or pathway variations associated with overall survival, sidedness, RAS mutation status and synchronous disease.

Materials and methods

Study population

Seventy-seven patients with mCRC in whom WES (Whole Exome Sequencing) analysis was performed in 2018 as part of routine care, and interpreted according to the Molecular Tumor Board of the Georges François Leclerc Cancer Center, were included in this single-centre, retrospective study. WES analyses were performed during first or second line therapy. WES analysis is performed as part of routine care in our center in order to find potential targetable mutations for second line therapy. Before patients consented to WES of their tumoral tissue, they were informed by their oncologist. Germline testing was performed after counseling by a clinical geneticist.

Only patients from whom informed consent was obtained and recorded in the medical chart were included in this retrospective study. The study was approved by the CNIL (French national commission for data privacy) and the local ethics committee, and was performed in accordance with the Helsinki Declaration and European legislation.

Sample selection

Physicians selected an archival tumor sample (primary or metastasis) for genomic analysis. At the discretion of the physician, a new tumor biopsy could be proposed to the patient. Tumor cellularity was assessed by a senior pathologist on a hematoxylin and eosin slide from the same biopsy core used for nucleic acid extraction and molecular analysis.

DNA isolation

DNA was isolated from archival tumor tissue using the Maxwell 16 FFPE Plus LEV DNA purification kit (Promega Corp.). DNA from whole blood (germline DNA) was isolated using the Maxwell 16 Blood DNA Purification kit (Promega Corp.) following the manufacturer's instructions. The quantity of extracted genomic DNA was assessed by a fluorometric method with a Qubit device.

Whole exome capture and sequencing

Two hundred ng of genomic DNA were used for library preparation, using the Agilent SureSelectXT reagent kit (Catalog number G9642B, Agilent Technologies, Inc.)and the All Exon v5 probeset (5190-8863, Agilent Technologies, Inc.). Following hybridization, the libraries were purified according to the manufacturer's recommendations and amplified by polymerase chain reaction (12 cycles). DNA integrity was verified using TapeStation (SCREENTAPE D1000 tapestation 5067-5582, reagents D1000 tapestation 5067-5583). Concentrations were measured using Qubit® dsDNA BR Assay Q32853. Loading concentrations were 22 nM for fragmentation and 6 pM for NextSeq injection. Normalized libraries were pooled, and DNA was sequenced on an Illumina NextSeq500 device using 2x111-bp paired-end reads and multiplexed. Names, catalog numbers and suppliers of the Illumina sequencing kit were following: NextSeq 500 High Output Kit FC404-2004/2140817 and NextSeq 500 Mid Output Kit FC404-2003/2140816. More than 90% of the target sequence was covered with a read depth of at least 10X for somatic DNA.

Exome analysis pipeline

Reads in FASTQ format were aligned to the reference human genome GRCh37 using the Burrows-Wheeler aligner (BWA v.0.7.15). Local realignment was performed using the Genome Analysis Toolkit (GATK v.3.6). Duplicate reads were removed using Picard v.2.5. To identify somatic single-nucleotide variants (SNVs), a validated pipeline was used that integrates mutation calls from three different mutation callers. Single Nucleotide Variants (SNVs) were called with VarScan (v2.4.3) (11) and Mutect (v1.1.7) (12) insertion/deletions (indels) were called with VarScan and Strelka (v2.9.2) (13). Tumor Mutational Burden (TMB) was calculated using the number of significant SNV (UTRs, synonyms, introns and intergenic SNVs filtered out) divided by the number of megabases covered at a defined level. TMB was calculated with and without splicing sites mutations. Splicing site mutations were excluded because it has been demonstrated that variants present in splice regions have predominantly no impact (14). To identify tumor-specific mutant peptides, pVAC-Seq v4.0.3(15) (personalized Variant Antigens by Cancer Sequencing) was used; pVAC-Seq is based on HLA typing obtained by HLAminer (16). TITAN (17) was used to infer the number of copy number alterations (CNA) subclones, the number of large deletions, as well as loss of heterogeneity (LOH)>15 Mb from whole-exome sequencing data. It was also used to estimate tumor ploidy. SNV signatures were generated using DeconstructSigs (v1.8.0) (18) and COSMIC signatures identified by Alexandrov et al (19). CNV signatures were inferred according to the methodology of Macintyre et al (20). MSI score was computed using MSIsensor (21) HRD score was obtained through scarHRD (22) pipeline.

Statistical analysis

Patient and disease characteristics were compared across the different groups of interest using the Chi-2 or Fisher's exact test for qualitative variables and the Wilcoxon test for continuous variables, as appropriate. Enrichr analysis using KEGG database was performed on genes differentially mutated given sidedness, metastases and KRAS mutation status (23). Genes with a P-value <0.1 were selected for this analysis. Enrichr is a web-based tool for analysing gene sets; it returns any enrichment of common annotated biological features, here KEGG database.

Survival analysis was performed using the survival R library. Continuous variables were dichotomised using Lausen et al (24) methodology through the maxstat library (25). The prognostic value of the different variables was tested using univariate Cox regression for overall (OS) survival. OS was defined as the time from diagnosis to death from mCRC. Survivors were censored at the end of study. Survival probabilities were estimated using the Kaplan-Meier method and survival curves were compared using the log-rank test.

Statistical analyses were performed using the R software (http://www.R-project.org/) and graphs were drawn using GraphPad Prism version 7.03 (GraphPad Software, LLC).

Results

Patients' clinical characteristics

We included 27 (35%) patients with right primary colon cancer and 50 (65%) with rectal or left primary colon cancer. Twenty-seven (35%) patients had metachronous metastasis and 50 (65%) were synchronous. Forty-six (60%) patients were RAS mutated. The most frequent metastasis site was the liver (54 patients, 70%). Liver metastasis were more frequently synchronous (44 patients, 57%), than metachronous (10 patients, 12%) (P-value=0.01). The most common sites of first metastasis were the liver (74%) and the lung (27%), two metastatic sites that are potentially curable by resection. The other identified sites of metastases were the lymph nodes, peritoneal, adrenal and bone metastasis.

The presence of lung or liver metastases at time of diagnosis of metastatic disease did not vary significantly by primary tumour site (19% of right-sided tumors versus 32% of left-sided mCRC for lung metastasis, and 18% of right-sided tumors versus 36% of left-sided mCRC for liver metastasis). In contrast, peritoneal and omental metastases were more frequent among right-sided primary tumors (P-value=0.001). The main differences between clinical variables according to tumor sidedness are presented in Table SI, according to synchronous or metachronous status in Table SII, and between RAS/WT mutations in Table SIII.

All patients received a doublet or triplet of chemotherapy as first-line therapy; fluorouracil and folinic acid were used consistently. Patients with synchronous metastasis significantly more frequently received oxaliplatin in their chemotherapy regimens (P-value=0.01). Thirty patients were treated with anti-epidermal growth factor receptor (EGFR), without any significant difference concerning the sidedness or the time to metastasis (P-values 0.13 and 0.61 respectively). Sixty-four patients (83%) were treated with anti-angiogenesis drugs (bevacizumab or aflibercept). Fifty-four patients (70%) received a therapeutic proposition from the molecular tumor board. Only 18 patients (23%) received a treatment based on the molecular tumour board recommendations, most frequently using an oral MEK-inhibitor called Trametinib (Table I). Three patients yielded significant clinical benefit from these therapeutic strategies, with more than 6 months of progression-free survival. Two of these three patients had unstable microsatellite status and were treated with immunotherapy.

Table I Description of treatment based on molecular tumour board recommendation.

Table I

Description of treatment based on molecular tumour board recommendation.

Sample	Somatic mutation	Nucleotide variant	Protein variant	Impact	Treatment
1	BARD1, BRIP1	c.266C>T, c.2665C>A	p.Pro89Leu, p.Gln889Lys	Unknown	OLAPARIB
2	NF1	c.1007G>A	p.Trp336Ter	Unknown	TRAMETINIB
3	BRAF	c.1799T>A	p.Val600Glu	Activating function	VEMURAFENIB
4	RAD51C	c.859A>G	p.Thr287Ala	Loss of function	OLAPARIB
5	MSH6	c.2017C>A	p.Pro673Thr	Unknown	NIVOLUMAB
6	TOP1	c.852G>A	p.Lys284Lys	Unknown	IRINOTECAN
7	TP53	c.450_451delAC	p.Pro152AlafsTer28	Loss of function	AFLIBERCEPT
8	KRAS	c.35G>C	p.Gly12Ala	Activating function	TRAMETINIB
9	MSH6	.3254delC	p.Phe1088SerfsTer2	Loss of function	DURVAUMAB
10	TP53	c.638G>T	p.Arg213Leu	Loss of function	BEVACIZUMAB
11	MSH6	c.2731C>T	p.Arg911Ter	Unknown	NIVOLUMAB
12	High number of variant	-	-	-	NIVOLUMAB
13	KRAS	c.35G>T	p.Gly12Val	Activating function	TRAMETINIB
14	MTOR	c.6352C>T	p.Leu2118Phe	Unknown	EVEROLIMUS
15	TP53	c.637C>T	p.Arg213X	Loss of function	REGORAFENIB
16	KRAS	c.38G>A	p.Gly13Asp	Activating function	TRAMETINIB

Patients' genomic characteristics

The most frequent mutation in the whole cohort was APC followed by TP53 and RAS (Fig. 1A). A summary of the genomic characteristics is presented in Table II. Thirty one patients presented WT mutational status, 44 patients presented KRAS and 2 patients NRAS mutational status. We identified 8 patients (10%) with BRAF mutation. The most frequent RAS mutation was KRAS (60%).

Figure 1 Genetic characteristics in the whole cohort. (A) Donut representing the proportion of patients presenting each of the top 10 mutated genes. (B) Boxplots representing the distribution of genomic characteristics that were significantly different between MSS and MSI patients, namely TMB score with and without splice regions, number of neopeptides according to whether they present weak or strong affinity, and HRD score. P-values were obtained using Wilcoxon tests. TMB, tumor mutational burden; MSS, microsatellite stability; MSI, microsatellite instability; HRD, homologous recombination deficiency.

Table II Summary of genomic characteristics by primary side, metachronous/synchronous and RAS/WT status.

Table II

Summary of genomic characteristics by primary side, metachronous/synchronous and RAS/WT status.

		Primary side			Moment of metastasis diagnosis			RAS status
Variable	Whole population (n=77)	Right side (n=27)	Left side (n=50)	P-value	Metachronous group (n=27)	Synchronous group (n=50)	P-value	RAS mutated (n=46)	RAS WT (n=31)	P-value
MSI categories, n (%)				>0.99			0.46			>0.99
MSS (score ≤5)	72 (93.5)	25(92)	47(94)		24(89)	48(98)		44(96)	29(94)
MSI low and high	5 (6.5)	2(8)	3(6)		3(11)	2(2)		2(4)	2(6)
TMB without splicing categories, n (%)				0.41			0.66			>0.99
Low (score ≤40)	71(92)	26(96)	45(90)		24(89)	47(94)		42(91)	29(94)
High	6(8)	1(4)	5(10)		3(11)	3(6)		4(9)	2 6)
RAS mutation, n (%)	46(60)	18(69)	28(56)	0.65	17(63)	29(58)	0.46	46(100)	-	-
BRAF mutation, n (%)	8(10)	5(19)	3(6)	0.18	5(19)	3(6)	0.12	-	8(10)	>0.99
MSI score, median (IQR)	0.08 (0.25)	0.03 (0.27)	0.09 (0.22)	0.35	0.23 (0.43)	0.09 (0.22)	0.35	0.06 (0.21)	0.10 (0.24)	0.40
TMB without splicing regions score, median (IQR)	5.5 (3.1)	5.7 (2.8)	5.4 (2.9)	0.55	5.7 (2.8)	5.4 (2.9)	0.5	5.6 (3.1)	5.4 (2.7)	0.7
HRD score, median (IQR)	23(13)	23(14)	23. (11.7)	0.87	23(14)	23(12)	0.87	21(13)	28(14)	0.07
Ploidy, median (IQR)	2.1 (0.5)	2.1 (0.5)	2.1 (0.5)	0.73	2.1 (0.4)	2.1 (0.5)	0.73	2.1 (0.2)	2.1 (0.6)	0.3
Clonality, n (%)				0.06			0.06	2 (1.75)	1 (1.75)	0.3
1	39(51)	9(33)	30(60)		9(33)	30(60)		-	-
>1	37(49)	17(63)	20(40)		17(63)	20(40)		-	-
Deletions with LOH, median (IQR)	3(8)	1(7)	4(9)	0.21	1(7)	4(9)	0.21	5(9)	1(5)	0.1
Microdeletions, median (IQR)	3(4)	3 (3.)	3(5)	0.64	3(3)	3(5)	0.64	3(5)	2(4)	0.1
Neoantigens, median (IQR)	11(8)	11(8)	11(7)	0.7	11(8)	11(7)	0.88	11(6)	11(10)	0.9
Strong neoantigens, median (IQR)	1(2)	1(1)	1(2)	0.07	1(2)	1(2)	0.63	1(2)	1(2)	0.9

[i] MSI, microsatellite instability; MSS, microsatellite stability; TMB, tumor mutational burden; HRD, homologous recombination deficiency; LOH, loss of heterozygosity; IQR, interquartile range; WT, wild-type.

Mean tumor mutational burden (TMB) without splicing regions in the whole cohort was 8.87 (median=5.50, IQR=3.07). The mean number of neoantigens was 13 (median=11, IQR=7.75). Five patients had a high HRD score, using the classical cut off value 42(26). The two most frequent SNV signatures were 1 and 25. The two most frequent CNV signatures were 3 and 5.

Seventy-two (94%) patients had proficient mismatch repair status (pMMR) and 5 (6%) had deficient mismatch repair status (dMMR) on immunohistological analysis (Table SIV). Median MSI score and TMB score were significantly higher in the dMMR group (respectively P-value=0.02 and P-value=0.01) (Fig. 1B). There was no significant enrichment of signature 6 in patients with dMMR status (P-value=0.7).

Association of genomic variables with sidedness

To evaluate whether we could isolate a genetic basis for the difference in survival between disease that originates in the right versus left side of the colon, we analysed genomic structural and gene alterations for the primary tumor site. The only difference in structural variants was significant enrichment of signature 26 in left-sided primary tumors (Fig. 2A). Signature 26 is associated with homologous repair deficiency (defective DNA mismatch repair).

Figure 2 Genetic characteristics discriminating tumor sidedness. (A) Distribution of the proportion of single nucleotide variation signature 26 according to primary tumor side. (B) Proportion of patients presenting at least one mutation in genes for which Fisher tests were considered significant (P<0.05), between primary tumor sides, by group and in the whole cohort. (C) Distribution of patients presenting an alteration in significant signaling pathways by primary tumor side. Only pathways for which the Fisher tests were considered significant (P<0.05) are presented. RTK, receptor tyrosine kinase; WT, wild-type.

For gene base analysis (Fig. 2B), there was significant enrichment of oncogenic alterations in AMER1, SMARCB1, ERBB4, FUBP1 and PTEN in right-sided primary tumors. Nevertheless, we did not identify any significant differences according to sidedness or frequencies of KRAS/BRAF mutations. Mutations observed in these genes are further described in Table SV. Results of Enrichr analysis using KEGG database are described in Table SVI.

Beyond the gene-level associations, analysis at the level of oncogenic pathways showed that mutations related to certain pathways differed according to primary tumor site. These pathways consisted in significant enrichment of Nucleotide excision repair for right-sided tumors, and RTK for left-sided ones (Fig. 2C). No significant association was found between sidedness and TP53/ATM, WNT/CTNNB1, TGF-beta or IGF2/PI-3-kinase pathways, which are frequently mutated in CRC.

Association of genomic variables with synchronous versus metachronous presentation

Using structural genome analysis, SNV signature 3, which is associated with failure of DNA double-strand break repair by homologous recombination, was significantly associated with synchronous presentation (Fig. 3A). We did not find additional differences at this genomic level.

Figure 3 Genetic characteristics discriminating synchronous or metachronous disease. (A) Distribution of the proportion of SNV signature 3 according to synchronous or metachronous status. (B) Proportion of patients presenting at least one mutation in genes for which the Fisher tests were considered significant (P<0.05), between synchronous or metachronous status, by group and in the whole cohort. (C) Distribution of patients presenting an alteration in significant signaling pathways by synchronous or metachronous status. Only pathways for which the Fisher tests were considered significant (P<0.05) are presented. WT, wild-type.

P53 loss of function mutations were significantly more frequent in synchronous metastasis. In contrast, metachronous metastasis was associated with PTCH, PTPRD and CSF1R mutations (Fig. 3B). Mutations observed in these genes are further described in Table SV. Results of Enrichr analysis using KEGG database are described in Table SVI.

When pooling mutations in gene pathways, we observed that Mismatch repair and Hedgehog signaling were significantly more frequently affected in metachronous tumors (Fig. 3C).

Association of genomic variables with RAS status

In our series, the prevalence of RAS mutations was not significantly different according to tumor location. On gene variation, KMT2B and RET were significantly associated with RAS wild type tumors. In contrast, KMT2C and GNAS were significantly more frequent with RAS mutated tumors (Fig. S1A). Mutations observed in these genes are further described in Table SV. Results of Enrichr analysis using KEGG database are described in Table SVI.

An analysis at the level of oncogenic pathways demonstrated that ERK signaling and chromatin organisation pathways were significantly associated with RAS mutation, while Phosphatidylinositol signaling was associated with WT tumors (Fig. S1B).

Association of genomic variables with survival

Median OS was 3.75 years. Regarding genomic structural variants, low SNV signature 15 was associated with poor overall survival (Fig. 4A). In gene variation, PDGFRA, SMAD4, PTEN, MYCN mutations were significantly associated with poor prognosis (Fig. S2A-D). We also estimated a multivariate model involving the 4 genes; all genes remain significant. At the level of oncogenic pathways, cell-cycle, NFKappaB signaling and mismatch repair pathways were significantly associated with better survival (Fig. 4B-D). We did not observe a significant association between any WNT-signaling pathway and CRC survival.

Figure 4 Genomic variables and overall survival. Kaplan Meier curves for variables with significantly different overall survival according to: (A) SNV signature 15, (B) oncogenic pathways NF Kappa B, (C) cell cycle, (D) mismatch repair, (E) synchronous status and (F) combination of mismatch repair and synchronous status. HR, hazard ratio; MMR, mismatch repair; SNV, single nucleotide variation; WT, wild-type.

Overall survival was significantly better in patients with metachronous metastatic disease, but this difference was no longer significant after removing MMR deficient tumors, thus suggesting a link between the prognosis of metachronous disease and enrichment in MMR deficient tumors (Fig. 4E and F).

Discussion

CRC is caused by multiple risk factors, including environmental, lifestyle and genetic risks. All these elements cause mutations and epigenetic alterations, conferring on cells the capacity to transform and grow, with aberrant DNA editing and defective DNA maintenance.

Next-generation sequencing (NGS) has identified a diversity of driver mutations in genes and altered signaling pathways in CRC (10,27). In addition, genomic structural patterns within gene pathway mutations could be used to identify prognostic features. Recent studies have revealed a mutational landscape of colorectal cancer and defined different subtypes that could guide therapeutic decisions (28). Increasing access to WES and the increasing rapidity of analysis offers new opportunities to implement such tests in the care of mCRC. WES analysis is performed in routine care in our center in order to find potentially targetable mutations for second line therapy.

Like in previous studies of precision medicine, we found that only very few mCRC patients yielded a benefit from precision medicine. In our study, we used WES analysis to identify genomic mutational profiles linked to tumor sidedness and metastatic occurrences, as well as genomic features related to prognosis.

In a systemic review and metanalysis (8), left-sided primary tumors were associated with improved prognosis, for both localized and metastatic tumors, in comparison to right-sided disease. Peritoneal and omental metastases, which are metastatic sites known to be associated with poor survival (29), were more frequent among right-sided primary tumors. Moreover, the survival differences seen between patients with right-sided vs. left-sided primary tumor sites in mCRC are supported by differences in transcriptomic patterns (8,30).

Like other authors (30,31), we showed that metachronous metastatic status is associated with better outcome. However, the molecular mechanism remains obscure, and the prognostic role seems to be less impactful than that of MMR status.

Our findings indicate that differences in survival could also be explained by genomic differences. Our analysis relied on structural information and on mutated genes grouped in pathways. We showed in particular that SNV signature 26, associated with MMR deficiency 29 (https://cancer.sanger.ac.uk/cosmic/signatures_v2.tt), was surprisingly associated with left-sided mCRC, while SNV signature 3 was associated with synchronous metastasis. Signature 3 is associated with better overall survival in ovarian (32) and breast cancer (33), by the better response to platinum therapy. At present, no data link signature 3 to oxaliplatin efficacy in mCRC. This question may be relevant for further clinical trials.

Moreover, somatic mutations in TP53 genes, which are present in the majority of cancers, are associated with poorer clinical outcomes, in several cancer types, including CRC (28). P53 loss of function mutations were significantly more frequent in synchronous metastasis.

In contrast, when we looked at more classical genomic features such as tumor mutational burden, there was no significant difference between tumors according to sidedness or time to metastasis occurrence. Conversely, we observed that MMR deficiency was associated with better outcome. MMR deficiency is more frequent in metachronous tumors and this genetic event may explain the better prognosis of these patients.

Oncogenic alterations-KRAS, NRAS and BRAF, which exhibit resistance to EGFR therapy with panitumumab and cetuximab, did not differ in our cohort, as previously reported (34), according to sidedness or metastatic occurrence, thus suggesting that these parameters are probably weak prognostic factors. The RAS pathway is a cell signaling pathway that plays a key role in regulation of cellular proliferation, apoptosis, cellular differentiation and migration and angiogenesis. This pathway is often dysregulated in mCRC. The association between RAS mutations and activation of the ERK signaling pathway is already well known.

When we looked at classical somatic mutations in mCRC, we observed that APC was frequently mutated in our series. However, APC was not associated with survival. The WNT signaling pathway is also frequently mutated in CRC and we did not observe a significant association between any WNT-signaling pathway and CRC survival. SMAD4 loss was found to be associated with poor outcome in our series. Similarly, a recent report also showed an association between SMAD4 loss and poor CRC survival, resistance to chemotherapy and decreased tumor immune infiltration (35).

Both PDGF and PDGFR families play an important role in colorectal carcinogenesis, and PDGFR is frequently overexpressed in CRC. Activation of this pathway is frequently related to angiogenesis, invasion, metastasis and poor survival (36). In our study, mutations in PDGFRA were associated with poor prognosis.

The main limitation of our work is the low number of patients included, which may impact the statistical significance, and precluded multivariate analysis. Moreover, WES was made during first or second line of treatment, so mutation profiles were generated under therapeutic pressure, which changed the cancer characterizations. Unfortunately, we do not dispose of samples before treatments as a baseline. In addition, considering the sequencing result of PTEN, PDGFRA, MYCN or SMAD4, the lack of PCR verification step is a limitation of this study and should be done in future works. Thus, our results should be considered as descriptive and exploratory, and warrant confirmation in further studies including larger sample sizes.

In conclusion, with the development of targeted therapies, it seems necessary to be able to rapidly identify the molecular status of mCRC tumors. Large panel or exome sequencing is slightly effective in improving patient care with precision medicine; however, such analyses could reveal structural and pathway genomic features that are associated with sidedness, synchronous status, and overall survival. Although the cost of NGS is steadily declining, WES nonetheless remains expensive, and currently, this is one of the major limitations on the routine use of exome sequencing. Moreover, the number of tumour mutations that could be used to treat a given patient is limited in conventional clinical practice or clinical trials, and it is difficult to determine which patients will clinically benefit from exome sequencing. In all likelihood, before being used in routine practice, exome sequencing will be reserved for patients with advanced mCRC, after one or more lines of treatment, or for patients with very poor outcomes, who fail to respond to classical targeted therapy and chemotherapy. Indeed, the knowledge of the molecular status could lead to inclusion in therapeutic clinical trials with a direct benefit for our patients. More generally, patients with right-sided mCRC, synchronous metastasis or peritoneal and omental metastasis may benefit the most from such analysis. This important information could be used to improve patient stratification for clinical trials, and will lead to a new molecular classification of patients that could be helpful to finetune the future of mCRC therapy.

Supplementary Material

Genetic characteristics according to RAS status. (A) Proportion of patients presenting at least one mutation in the most frequently mutated genes, by RAS status. (B) Distribution of patients presenting an alteration in significant signaling pathways according to RAS status. Only pathways for which the Fisher tests were considered significant (P<0.05) are presented. WT, wild-type.

Gene mutation and overall survival. Kaplan Meier curves for genes with significantly different overall survival according to their mutation status, namely (A) PDGFRA, (B) SMAD4, (C) PTEN and (D) MYCN. HR, hazard ratio; WT, wild-type.

Summary of clinical characteristics according to primary tumour side (right/left).

Summary of clinical characteristics according to metachronous/synchronous status.

Summary of clinical characteristics according to RAS mutational status.

Description of genomic characteristics according to MSI/MSS groups.

Description of mutations observed in genes differentially mutated in function of sidedness, synchronous or metachronous status, and KRAS mutated status.

Results of Enrichr analysis based on Kyoto Encyclopedia of Genes and Genomes pathways and genes differentially mutated in function of sidedness, synchronous or metachronous status, and KRAS mutated status.

Acknowledgements

The authors would like to thank Ms. Sandy Chevrier (Department of Medical Oncology, Georges François Leclerc Cancer Center-UNICANCER, Dijon, France), Dr Romain Boidot (Department of Medical Oncology, Georges François Leclerc Cancer Center-UNICANCER, Dijon, France) for providing whole exome sequencing data, Mr. Hugo Mananet (Department of Medical Oncology, Georges François Leclerc Cancer Center-UNICANCER, Dijon, France) for pre-processing data, and Dr Juliette Albuisson (Department of Medical Oncology, Georges François Leclerc Cancer Center-UNICANCER, Dijon, France) for her valuable comments. The authors would also like to thank Dr Fiona Ecarnot (EA3920, University of Franche-Comté, Besançon, France) for English correction and helpful comments.

Funding

No funding was received.

Availability of data and materials

The datasets generated and analyzed during the current study are available in the Sequence Read Archive repository, (https://www.ncbi.nlm.nih.gov/sra/?term=SRP318854).

Authors' contributions

MDGDA, LG, ZT, CT and FG contributed to the design and implementation of the research, to the analysis of the results and to the writing of the manuscript. CT and FG confirm the authenticity of all the raw data. All authors read and approved the final manuscript.

Ethics approval and consent to participate

The present study was approved by the CNIL (French national commission for data privacy) and the Georges François Leclerc Cancer Center (Dijon, France) local ethics committee (13.085), and was performed in accordance with the Helsinki Declaration and European legislation. Written informed consent was obtained from all subjects involved in the study. Only patients from whom informed consent was obtained and recorded in the medical chart were included in this retrospective study.

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

References