Genomic mutation characteristics and prognosis of biliary tract cancer

Abstract

Background: The incidence of biliary system cancer is higher in the Chinese population than in the West. The overall prognosis of gallbladder cancer and cholangiocarcinoma is poor, and the current treatment is limited. In order to explore the pathogenesis of biliary tract cancers and potential targeted therapies, we mapped the mutation landscape of biliary tract cancer in the Chinese population and analyzed the molecular mechanism related to prognosis. Methods: A total of 59 formalin fixed paraffin-embedded (FFPE) tissue samples were obtained from patients with operable biliary tract cancer. We conducted targeted capture sequencing of 620 genes through high-throughput sequencing technology and analyzed the fusion information of 13 genes. Results: Mutations were detected in 88% samples, and the most frequent mutation base was C>T. Genes with higher single nucleotide variations (SNV) and copy number variations (CNV) frequency are TP53, KRAS, ARID1A, VEGFA, cyclin family related genes and cyclin-dependent kinase genes. Actionable mutations were detected in 59.3% samples, and germline mutations were detected in 22% samples. Patients with KRAS mutations, VEGFA pathway mutations and higher tumor mutation burden (TMB) may have poor prognosis. Conclusions: We explored the mutation characteristics and prognostic mechanism of biliary tract cancers in the Chinese population. This study provides potential evidence for targeted therapy and immunotherapy of biliary tract cancers.

Introduction

Biliary Tract Cancer (BTC) is a malignant tumor originating from the gallbladder and bile duct epithelium, including gallbladder cancer (GBC), intrahepatic cholangiocarcinoma (ICC) and extrahepatic cholangiocarcinoma (ECC). ECC consists of hilar cholangiocarcinoma and distal cholangiocarcinoma [1]. In Western countries, the incidence of cholangiocarcinoma is low, with an annual incidence of 0.35-2 per 100,000 [2]. In Asian countries, the incidence of cholangiocarcinoma is several times higher than that in Western countries. Gallbladder cancer is closely related to cholelithiasis and chronic cholecystitis, accounting for more than 80% of biliary system cancers. The global incidence of gallbladder cancer is 220,000, and the number of deaths is about 165,000 in 2018. In China, the incidence and mortality of gallbladder cancer are 52,800 and 40,700, and the morbidity and mortality are ranked 19th and 12th, respectively [3,4].

Regardless of location, BTC has potentially high transfer and invasion ability. Due to its anatomical location and distribution along the bile duct, it is difficult to remove completely by surgical resection. Even if it is diagnosed at an early stage, BTC is associated with poor prognosis [5,6]. Unfortunately, most BTC patients are diagnosed with advanced disease at their first visit and cannot be treated surgically. The 5-year survival rate of these patients is extremely low, about 10% for cholangiocarcinoma and less than 5% for gallbladder cancer [7-9].

There are no molecular markers related to clinical diagnosis, but drugs targeted to IDH1 and FGFR1/2/3 have shown superior performance recently [10-12]. The phase III clinical trial ClarlDHy demonstrated IDH1 inhibitor Ivosidenib improved the medium progression free survival (Ivosidenib 2.7 months vs. placebo 1.4 months) of patients with ICC who have been treated by chemotherapy [13]. FGFR2 target drugs such as infigratinib, derazantinib and TAS-120 showed good efficacy and controllable toxicity in phase II study [10-12]. The disease control rate of Infigratinib, a tyrosine kinase inhibitor, as a second-line drug for FGFR2 fusion-positive patients with advanced cholangiocarcinoma was 75.4% [12]. Irreversible pan-FGFR inhibitor, TAS-120, inhibited secondary mutations of FGFR2 and have efficacy in four patients with FGFR2-fusion-positive ICC who developed resistance to BGJ398 or Debio1347 [10]. On April 17th, 2020, the U.S. Food and Drug Administration (FDA) approved pemigatinib for FGFR2 fused cholangiocarcinoma and grants priority review to a variety of FGFR inhibitors for their superior efficacy. The pathogenesis of BTC is still unclear. Several previous studies [14-16] have presented common gene mutation spectrums in BTC in Japanese and western populations. Few studies outline genomic mutation characteristics of such tumors in the Chinese population. The next-generation sequencing with high throughput and high efficiency can perform parallel detection on multiple genes, which has been widely used in recent years. In order to explore the molecular mechanism of BTC and the population that would benefit from targeted therapy, we sequenced 620 genes related to tumorigenesis in 59 BTC samples.

Materials and methods

Clinical samples

We collected 59 surgical tissue samples of BTC from Eastern Hepatobiliary Surgery Hospital of the Second Military Medical University (Shanghai, China). According to WHO 2015 classification criteria, the tumor samples including 17 cases of gallbladder cancer and 42 cases of cholangiocarcinoma. All tissue samples were paired with blood samples to rule out nonpathogenic germline mutations. Follow-up data of 26 samples were obtained from Eastern Hepatobiliary Surgery Hospital of the Second Military Medical University. The clinical features included sex, age, tumor location, TNM stage, ECOG score and other information that are presented in Table 1. All the patients in our study signed an informed consent.

Table 1.

Characteristics of patients with BTC

Characteristic	Total (n=59)	Any mutation (n=52, 88.1%)	No mutation (n=7, 11.9%)
Age, median	57.2 (34-75)	56	65
Male gender	33	29 (87.9)	4 (12.1)
Female gender	26	23 (88.5)	3 (11.5)
ECOG
0	9	8 (88.9)	1 (11.1)
1	26	23 (88.5)	3 (11.5)
2	12	10 (83.3)	2 (16.7)
Unknown	12	11 (91.7)	1 (8.3)
Cancer type
Cholangiocarcinoma	42	37 (88.1)	5 (11.9)
Gallbladder cancer	17	15 (88.2)	2 (11.8)
Stage
I	1	1 (100)	0 (0)
II	7	5 (71.4)	2 (28.6)
III	14	12 (85.8)	2 (14.2)
IV	21	21 (100)	0 (0)
Unknown	16	13 (81.3)	3 (18.7)

Library building and sequencing

DNA was extracted from FFPE-fixed tumor tissue or peripheral blood using QIAamp DNA FFPE Tissue Kit (Qiagen). NimbleGen SeqCap EZ choice capture panel was used to capture the coding region of 620 genes (Supplementary Table 1) and the splicing sites. The DNA libraries were built according to the procedure of KAPA Hyper Prep protocols (KAPA). The final libraries were sequenced on the Illumina Novoseq6000 (PE150) sequencer, and the original FASTQ file was obtained. The final libraries were sequenced by Illumina Novoseq6000 (PE150).

SNV and indel calling

Trimmomatic was used to filter the sequenced FASTQ files, and Burrows-Wheeler Aligner (BWA) was used to align the reads with the reference genome GRCh37 (hg19). Duplicates generated by PCR were removed by SAMtools. SNVs were called by Mutect2 with a paired workflow. ANNOVAR was used to annotate the variants. We filtered the obtained SNVs according to the following conditions: (1) base quality value ≥20; (2) mutation reads depth ≥4; (3) variant allele frequency ≥1%; (4) reads supporting variation <4 in normal, tumor abundance/normal abundance ≥8; (5) no strand bias (GATK parameter FS >60 for SNP and FS >200 for indel); (6) discard synonymous mutations; (7) variation not in the dbSNP database.

Copy number variation (CNV) analysis

When the dispersion is normal, we set the cutoff values for CNV as 1.5 and 0.5 copies. It is regarded as copy number amplification when the value is larger than 1.5, and it is regarded as copy number deletion when the value is smaller than 0.5.

TMB calculating

TMB is defined as the number of SNV mutations per megabase. We kept mutations with mutation frequency ≥5% and removed synonymous mutations.

Driver mutation analysis

Driver genes were identified using oncodriveCLUST software, which is based on mutation frequency. The loss-of-function (LoF) gene mutations in the coding region were used as background. Gain-of-function (GoF) gene mutations were analyzed as key points.

Germline variant calling

Filter germline variants were obtained by GATK according to the following conditions: (1) mutation depth ≥50; (2) variation frequency ≥30%; (3) discard synonymous mutations; (4) population frequency ≤1/1,000 in ExAC, 1,000 genome and other database; (5) according to the ClinVar database, we reserved the splicing, stop-gain, frameshift, or (likely) pathogenic variants.

Survival analysis

All the analysis and graphs were based on R software. The survfit and survdiff functions in R were used to generate Kaplan-Meier survival curves and calculate the P value of the log-Rank test. Genes with more than 4 mutations were included in the survival analysis.

Results

Mutation signature

Target capture sequencing was conducted in all 59 BTC samples, with an average sequencing depth of 2,500X. At least one mutation was detected in 88% of tumor samples. In total, we identified 853 somatic mutations, including 736 SNVs and 99 indels. Among all the mutations, there were 649 missense, 77 nonsense, 78 frameshift, and 7 splicing mutations. The overall SNV mutation rate was 55.66 muts/Mb, and the Indel mutation rate was 7.49 muts/Mb. The mutation rates of cholangiocarcinoma and gallbladder cancer were 30.25 muts/Mb and 25.41 muts/Mb, respectively. In this study, C>T was found to be the most common type of mutation, accounting for 62.4% of all SNVs (Figure 1A). Non-negative matrix factorization (NMF) was used to identify mutant signatures of BTC. This analysis identified two different signatures: signature A is characterized by (A/C/T/G) CG> (A/C/T/G) TG, and signature B is characterized by TC (A/C/T) <TG (A/C/T) and TC (A/T/C/G) <TT (A/T/C/G) (Figure 1A). We compared the identified signature with the COSMIC signature and found that signature A is similar to COSMIC signature SBS1 and SBS6 (cosine similarity is 0.66 and 0.68 respectively), and that signature B is similar to COSMIC SBS13 (cosine similarity is 0.78) (Figure 1B).

Figure 1.

Mutational signatures. A. Mutational signatures identified in BTC. Two mutational signatures were detected in 59 patients’ tumor samples with BTC. B. Identified signatures compared to the COSMIC signatures. The two mutational signatures detected in the 59 BTC samples were compared with the corresponding COSMIC signatures determined by cosine similarity: signature A is similar to SBS1 (cosine similarity =0.66) and SBS6 (cosine similarity =0.68); signature B is similar to SBS13 (cosine similarity =0.78).

We profiled the somatic mutation maps of BTC (Figure 2A). A total of 603 genes were mutated in all BTC samples, and most of the gene mutations appeared at least in one sample. We found 47% of all 59 patients harbored the TP53 mutation, significantly higher than previously reported (33%, 26%) [15,16]. The mutation frequencies of TP53 in cholangiocarcinoma and gallbladder cancer were 43% and 80%, respectively (Figure 2B, 2C), and both are higher than previously reported. TP53 gene mutations had no hot spot, but were mainly in the DNA binding domain (Figure 3). In previous reports, KRAS, with mutation frequency ranging from 5% to 18%, was the gene with the second most significant number of mutations after TP53 [15-17]. In this study, the KRAS mutation frequency was 19%, and the frequencies in cholangiocarcinoma and gallbladder cancer were 24% and 13%, respectively (Figure 2B, 2C). We used oncodriverCLUST to analyze the possible driver genes and found that KRAS and IDH1 mutations were driver mutations (Supplementary Table 3).

Figure 2.

Somatic SNV and indel mutations of BTC. A. Somatic mutation landscape of SNV and indel in patients with Biliary Tract Cancers. Waterfall plots showing the frequency and types of TOP 40 somatic SNV and indel mutations found in 52 BTCs with mutations. B. Somatic mutation landscape of SNV and indel in patients with cholangiocarcinoma. Waterfall plots showing the frequency and types of TOP 20 somatic SNV and indel mutations found in 37 cholangiocarcinoma with mutations. C. Somatic mutation landscape of SNV and indel in patents with gallbladder cancer. Waterfall plots showing the frequency and types of TOP 20 somatic SNV and indel mutations found in 15 gallbladder cancers with mutations. Each column represents one patients’ sample and each row represents a feature. From top to bottom, the name and the mutated frequency of each gene are given. A color key is at the bottom.

Figure 3.

Mutation sites of TP53. Distribution of spot mutations on full length TP53 gene with a bar code representing their frequency was presented. Twenty-eight of 59 BTC patients’ samples were identified with TP53 mutated spots. No hot spot was found, and mutations were mainly in the DNA binding domains. A color key is at the bottom.

In addition to the above genes, other frequently mutated genes in this study included ARID1A, KMT2C, ATM, BRCA2, PBRM1, SMARCA4 et al. We estimated that ARID1A was mutated in 17% of all patients in this study, slightly higher than that was recorded in the COSMIC database (12%). The proteins encoded by the ARID1A, ARID1B, PBRM1 and ARID2 genes are all parts of the large ATP-dependent chromatin remodeling complex SNF/SWI. ARID1A is the largest subunit in the SWI/SNF chromatin remodeling complex, which has activities of helicase and ATPase, and regulates transcription by changing the chromatin structure of specific genes [18].

Copy number variation (CNV)

CNV is one of the driver factors of carcinogenesis and can directly affect gene transcription and protein expression. In this study, CNV was calculated based on the relative coverage of tumor samples and normal samples. We used GISTIC2.0 to analyze statistically significant local amplifications or deletions, and found that CNV occurred in 35.6% of patient samples (Figure 4A). Frequently amplified or deleted genes included vascular endothelial growth factor A encoding gene VEGFA, cyclin family genes (CCND1, CCND2, CCND3), cyclin dependent kinase genes (CDK12, CDK6), cyclin dependent kinase inhibition genes (CDKN2A, CDKN2B), Erb-B2 receptor tyrosine kinase encoding gene ERBB2 and other genes such as MYC, MDM4, MDM2.

Figure 4.

Somatic copy number variation (CNV) spectrum of BTC. A. Waterfall plots showing the frequency and types of the TOP 28 somatic CNV found in BTC. Each column represents one of the 59 BTC samples and each row represents a feature. From top to bottom, the name and the mutated frequency of each gene are given. A color key is at the bottom. B. CNVs concentration region of 59 BTC patients on the chromosome.

Statistical analysis showed that the chromosome segments that are more prone to amplification and deletion were 6p21.1 and 5q35.3, respectively (Figure 4B). The 6p21.1 region occurs in a variety of cancer types in Chinese population, including lung cancer and esophageal cancer, and is also associated with various familial genetic diseases such as hypertension and atherosclerosis [19]. CNV in the 5q35.3 region is relatively rare and has been reported in pancreatic cancer [20].

Pathway enrichment analysis

In order to explore the molecular mechanism of BTC, KEGG pathway enrichment analysis was used to enrich somatic mutant genes into important signaling pathways. Among these pathways, the tumor associated signal pathways with frequent gene mutations are PI3K-Akt, MAPK and Ras signaling pathways (Figure 5). PI3K-AKT signaling pathway (Supplementary Figure 1A) not only regulates tumor cell proliferation, but also is closely related to tumor angiogenesis [21]. The Ras-Raf-MAPK signal transduction pathway (Supplementary Figure 1B) is involved in the signal transduction of various activated growth factors, cytokines, mitogens and hormone receptors, and plays an important role in regulating cell proliferation, growth and differentiation. Mitogen activated protein kinase (MAPK) can promote vascular endothelial cell proliferation and neovascularization, which can provide more nutrients for tumors, accelerate tumor growth, and promote the spread of cancer cells [22].

Figure 5.

Signal transduction pathway enrichment. Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were conducted using the clusterProfiler package of R software. P value <0.05 was set as the cutoff criterion. The tumor associated signal pathways with frequent gene mutations are PI3K-Akt, MAPK, and Ras signaling pathways.

Germline variants in BTC patients

In order to explore the genetic characteristics of cancer in the biliary system, we analyzed germline mutations in 61 genes (Supplementary Table 2) associated with genetic susceptibility. Germline mutations in BTC have been reported in several studies, and the reported detection frequency is 10%-20% [23]. Germline mutations were found in 13 (22%) patients’ samples in this study. Each germline mutation was detected in only one patient sample, and most of them were truncated, frameshift or splicing variations with high evidence of pathogenicity. Germline mutations are mostly DNA damage repair related genes, such as Lynch syndrome related mismatch repair genes MLH3, MSH6 [24,25], nucleotide excision repair gene ERCC4, DNA single-strand damage repair gene XRCC1. Germline mutations have also occurred in other genes including POLD1, XRCC1, IKZF1, ERCC4, TLR4, EPPK1, CDKN1A, NCOA3 and FGFR1. So far all these mutation nucleotide sites above have not been reported in biliary tract cancers (Table 2).

Table 2.

Germline gene mutations

Patient ID	Cancer Type	Gene	Chrom	Start_Position	End_Position	Nucleotide change	Protein change	Location	Variant classification
1810276	cholangiocarcinoma	IKZF1	chr7	50459525	50459525	c.688G>A	p.A230T	exon6	Missense
1810285	gallbladder carcinoma	TLR4	chr9	120476505	120476505	c.2099dupT	p.P701fs	exon3	Frame_Shift_Ins
1810285	gallbladder carcinoma	POLD1	chr19	50912119	50912119	c.1853dupA	p.Y618_T619delinsX	exon15	Nonsense
1810455	cholangiocarcinoma	EPPK1	chr8	144947313	144947313	c.108dupC	p.R37fs	exon2	Frame_Shift_Ins
1810558	gallbladder carcinoma	RFWD2	chr1	176054978	176054978	c.1075C>T	p.R359X	exon10	Nonsense
1830007	gallbladder carcinoma	XRCC1	chr19	44079097	44079097	c.109A>T	p.K37X	exon2	Nonsense
1830027	cholangiocarcinoma	CDKN1A	chr6	36652317	36652317	c.439A>G	p.M147V	exon2	Missense
1910201	cholangiocarcinoma	NCOA3	chr20	46279949	46279949	c.3875T>C	p.M1292T	exon20	Missense
1910783	cholangiocarcinoma	FGFR1	chr8	38271190	38271190	c.2518C>T	p.R840X	exon19	Nonsense_Mutation
1912566	cholangiocarcinoma	ERCC4	chr16	14026143	14026143	c.1102+1G>A	.	intron6/7	Splice_Site
1912694	cholangiocarcinoma	INSRR	chr1	156816322	156816322	c.1798dupA	p.T600fs	exon8	Frame_Shift_Ins
1912695	cholangiocarcinoma	TRAF3	chr14	103342048	103342048	c.385A>G	p.M129V	exon4	Translation_Start_Site
1930095	cholangiocarcinoma	MLH3	chr14	75513610	75513610	c.2749C>T	p.Q917X	exon2	Nonsense
1930184	cholangiocarcinoma	MSH6	chr2	48033987	48033987	c.4068_4071dupGATT	p.K1358fs	exon10	Frame_Shift_Ins

Survival impacts of molecular characteristics

To analyze the relationship between molecular characteristics and prognosis, we obtained survival information of 26 patients and plotted Kaplan-Meier survival curves by univariate analysis. Patients with KRAS mutations had worse median overall survival (OS, 108 d vs. 320.5 d, p=0.00057) (Figure 6A) among all patients in this study. We also compared the effects of mutations in various signaling pathways on survival and found that gene mutations in the VEGFR signal pathway had a significantly negative impact on OS (144.5 d vs. 324.5 d, p=0.0077) (Figure 6B). In addition, we analyzed the correlation between TMB and survival in stage IV patients. Patients whose TMB was higher than the median TMB had worse prognosis than those below (172.5 d vs. 474 d, p=0.05) (Figure 6C).

Figure 6.

Kaplan-Meier curves of patients with mutated and wild genes. Survival analyses were performed using the survival package of R software. Each gene was evaluated individually using the Kaplan-Meier method in R. A. Kaplan-Meier curve of patients with mutated (n=4) and wild (n=22) KRAS gene. B. Kaplan-Meier curve of patients with mutated (n=6) and wild (n=20) VEGF signaling pathway. C. Kaplan-Meier curve of patients with high (n=8) and low (n=5) TMB group.

Discussion

We identified some molecular characteristics of biliary tract cancers through targeted capture sequencing. Among all the mutated bases, C>T bases accounted for the largest proportion, which is consistent with previous reports [15-17]. The mutation signatures were similar to COSMIC signature SBS1, SBS6 and SBS13. Signature SBS1 is an endogenous mutation process, triggered by spontaneous or enzymatic deamination of 5-methylcytosine to thymine, resulting in G:T mismatch in double-stranded DNA. Failure to detect and eliminate these mismatches before DNA replication always causes C to replace T. Signature SBS6 is associated with DNA mismatch repair deficiency, and often occurs in microsatellite unstable tumors. SBS13 is related to the cytosine deaminase activity of the AID/APOBEC family which may be caused by the replication of the basic site generated by the error-prone polymerase (such as REV1) during base excision and repair of uracil [26]. Series of DNA damage and repair deficiency may be the internal cause of biliary tract cancers.

TP53 encodes a tumor suppressor protein which contains transcriptional activation, DNA binding and oligomeric domains. The protein encoded by TP53 responds to various cell pressures, regulates target gene expression, and thereby induces cell cycle arrest, apoptosis, aging, DNA repair or metabolic changes. TP53 mutations are ubiquitous in many cancer types, most of which are frameshift or nonsense mutations that lead to protein inactivation. These mutations are widely distributed throughout the whole gene, and occur most frequently in the DNA binding domain [27,28]. KRAS mutation is a key driver of oncogenesis [29]. We also found that frequently mutation of KRAS might be the driver mutation of BTC. Ongoing cancer driver gene discovery efforts have identified many new drivers within the RAS pathway [30,31]. In addition, there are already preliminary results of clinical trials, and AMG510 targeting KRAS G12C is particularly prominent [32]. It has opened a historic gap and helped patients with KRAS mutation. Adagrassib (mrtx849) has excellent clinical data recently [33]. Genes encoding SWI/SNF chromatin remodeling complexes such as ARID1A, ARID1B, PBRM1 and ARID2 were also frequently mutated in this study. Recent studies have found that ARID1A mutation is associated with microsatellite instability, C>T mutation pattern, and increased mutational burden in many kinds of tumors. Those tumors formed by ARID1A-deficient ovarian cancer cell lines in mice have increased mutations, more tumor-infiltrating lymphocytes and PD-L1 expression. PD-L1 monoclonal antibodies can reduce tumor burden and prolong survival of mice, but it cannot inhibit ARID1A wild-type ovarian tumors. ARID1A deficiency may lead to tumor MMR deficiency, and patients with ARID1A deficiency may benefit from immune checkpoint inhibitors [34].

Genes encoding vascular endothelial growth factor A (VEGFA), cyclin family, cyclin-dependent kinase, cyclin-dependent kinase inhibitor and others had higher frequency of CNV than others. The protein encoded by VEGFA belongs to the PDGF/VEGF family whose products play an important role in angiogenesis and endothelial cell growth. The expression of VEGFA is upregulated in various tumors including cholangiocarcinoma, which can induce endothelial cell proliferation, promote tumor cell migration and inhibit apoptosis [35-37]. Given the importance of CCND and CDK, the development of CDK4/CDK6 inhibitors has been a strategy for the generation of new anticancer drugs [38,39].

In this study, the incidence of germline mutations was 22%, slightly higher than previously reported [23]. Germline mutations of DDR genes are associated with increased mutation load, which causes patients with these mutations to possibly be more sensitive to PARP inhibitors and platinum-based treatments. Germline mutations may be a susceptible factor for the occurrence of BTC. Screening for germline mutations in susceptible people is of great importance in prevention and treatment of biliary tract cancers.

Patients with KRAS mutations and VEGF signaling pathway mutations have shorter overall survival. As a cancer-driven event, KRAS mutation is a predictor of drug resistance and poor prognosis for various cancers [29]. Abnormal VEGF signaling pathway promotes tumor cell proliferation and migration, often leading to poor prognosis [40,41]. In addition, TMB-H patients also exhibited poor OS in this study, possibly due to complicated tumor clone composition and carcinogenesis mechanism. In clinical practice, patients with poor prognosis are suggested to adopt more aggressive treatment and rigorous and regular follow up. Patients with high TMB may be eligible for immunotherapy such as PD-1/PD-L1 inhibitors.

According to the classification of oncoKB, the proportion of actionable gene mutations in the study was up to 59.3% (Supplementary Table 4). As more therapeutic targets are in research and more targeted drugs are being approved, there are more opportunities for biliary tract cancer patients with actionable gene mutations to receive precision medicine in the future. With the rapid development of the next-generation genome sequencing technology, defining of somatic or germline mutation in BTC patients could help accurately identify patients benefiting from drugs such as PARP inhibitors and immune checkpoint inhibitors.

The limited number of samples, insufficient clinical information and not using the whole exon sequencing may be deficiencies of this study. In later research, it is necessary to expand the research cohort with comprehensive clinical information and analyze samples with the whole exon sequencing and other necessary methods. Future prospective research is needed to verify the correlation between specific molecular characteristics and prognosis.

Disclosure of conflict of interest

None.