Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Somatic mutations in arachidonic acid metabolism pathway genes enhance oral cancer post-treatment disease-free survival

Abstract

The arachidonic acid metabolism (AAM) pathway promotes tumour progression. Chemical inhibitors of AAM pathway prolong post-treatment survival of cancer patients. Here we test whether non-synonymous somatic mutations in genes of this pathway, acting as natural inhibitors, increase post-treatment survival. We identify loss-of-function somatic mutations in 15 (18%) of 84 treatment-naïve oral cancer patients by whole-exome sequencing, which we map to genes of AAM pathway. Patients (n=53) who survived ≥12 months after surgery without recurrence have significantly (P=0.007) higher proportion (26% versus 3%) of mutations than those who did not (n=31). Patients with mutations have a significantly (P=0.003) longer median disease-free survival (24 months) than those without (13 months). Compared with the presence of a mutation, absence of any mutation increases the hazard ratio for death (11.3) significantly (P=0.018). The inferences are strengthened when we pool our data with The Cancer Genome Atlas (TCGA) data. In patients with AAM pathway mutations, some downstream pathways, such as the PI3K–Akt pathway, are downregulated.

Introduction

The arachidonic acid (AA) metabolism pathway, hereinafter referred to as the AAM pathway, a key inflammatory pathway involved in cellular signalling, has been implicated in diverse human cancers1. This pathway generates genotoxic agents, such as, reactive oxygen species, peroxides, free organic radicals and aldehydes that promote tumorigenesis2,3. AA, a common dietary fatty acid stored in cell membranes and in extracellular microenvironment, is released in response to extracellular stimuli and mobilized by phospholipase A2 (PLA2) and converted to prostaglandins, hydroperoxyeicosatetraenoic acids and epoxyeicosatrienoic acids by the action of cyclooxygenases (COXs), lipoxygenases (LOXs) and cytochrome P450 (CYP) epoxygenases, respectively1,4,5. AAM pathway promotes tumour progression by activating major mitogenic signalling pathways, including PI3K, MAPK and Jun kinase6,7,8. AA metabolism cascade is involved in head and neck squamous cell carcinoma (HNSCC)9,10,11,12, particularly in oral squamous cell carcinoma (OSCC)13.

OSCC is the eight most common cancer in the world and constitutes a major cancer burden predominantly among males in India14. In most (60%) Indian OSCC patients, the gingivo-buccal (GB) area—comprising the buccal mucosa, the retromolar trigone and the lower gum—is affected, while in Western patients the tongue is predominantly (65%) affected15,16. Most OSCC-GB patients present at advanced stages (stages III and IV), have a very high locoregional failure rate and their rate of mortality has remained high and unchanged for over three decades despite the best multimodal treatment17,18. Recently, we14 and others19,20,21,22 have identified, by exome sequencing, somatic driver mutations and key pathways in HNSCC, which show that OSCC is primarily driven by mutations in TP53, FAT1, CASP8, NOTCH1 and USP9X genes; these genes are tumour suppressors and, therefore, difficult to target by drugs. Several members of AAM pathway were reported to overexpress in different cancers and modulate cancer progression23,24. Median disease-free survival (DFS) in cholangiocarcinoma patients was significantly better in patients with low levels of 5-LOX or COX-2 expression than those with high levels25. Lipid autacoids that are CYP-derived metabolites of AA appear to be involved in cancer either by regulating cell-autonomous tumour survival and growth or by modulating stromal processes, such as angiogenesis and inflammation that can support tumour progression26. Recent studies12,27,28,29 have reported that inhibitors of different components of the AAM pathway show promising results in the treatment of various human cancers including premalignant lesions of HNSCC30. Synthetic inhibitors of CYP significantly reduced tumour size in intracranial glial tumours and prolonged survival of treated rats31. A clinical trial ( http://clinicaltrials.gov/ct2/show/NCT00620139) is underway to study the proximal effects of concurrent chemotherapy on the expression of potential molecular targets on the AAM pathways in HNSCC.

Since AAM pathway inhibitors hold promise for cancer treatment, including prolonging post-treatment survival25,31, we hypothesize that non-synonymous somatic mutations in genes of the AAM pathway act as natural inhibitors of the pathway and enhance post-treatment survival probability of OSCC-GB patients. We analyse the somatic mutation profile of genes on the AAM pathway in OSCC-GB patients and discover robust evidence in support of the hypothesis.

Results

Case description, human papillomavirus infection and exome sequencing

Tumour and blood samples from 110 patients diagnosed with OSCC-GB were collected, with informed consent, at the time of surgical excision. Three patients who died of causes other than cancer were excluded. Twenty-three living patients for whom the follow-up period was <12 months were also excluded. The remaining patients (n=84) included in this study (see Supplementary Table 1, for their characteristics) were categorized into two subgroups based on whether (SRV12; n=53) or not (N-SRV12; n=31) they survived for at least 12 months after surgery without recurrence. (Varying the period of survival by ±6 months to define the subgroups did not change inferences; results not shown.) Survival and recurrence data were collected by following up the treated patients from the date of surgery to 39 months—the longest observed duration of survival. Median DFS was 14 months. In the total cohort, the percentage of patients who were alive for >1 year was 63%, >2 years was 33% and >3 years was 2%. Recurrence at the primary site was confirmed in 40 out of 84 patients (47.6%). Human papillomavirus infection (HPV) infection, all with high-risk subtypes (16 and 18), was detected in 20% of patients. Tumour-normal pairs of the 84 patients were sequenced at a mean depth of 44.94±1.83 (details in Supplementary Table 2). In addition, exome sequence data were also generated for 17 of the 23 patients who were alive for a period of <12 months (short follow-up cases); their mutation profiles are given in Supplementary Table 3. Molecular data of these patients were not included in the survival analysis.

Longer OSCC-GB survivors have larger number of AAM mutations

Loss-of-function somatic mutations in the coding regions of one or more of the 63 AAM pathway genes (KEGG pathway id—hsa00590) were detected in 15 (18%) of the 84 tumours (Table 1). A significantly (P=0.007) higher proportion (26%) of patients who survived for >12 months without recurrence (SRV12 subgroup) possessed somatic mutations in one or more genes compared with those (N-SRV12 subgroup) who died within 12 months after surgery (3%). The median duration—24 months—of post-treatment survival of patients in the SRV12 subgroup was significantly (P<0.001) higher than patients in the N-SRV12 subgroup for whom the median duration was 6 months.

Table 1 List of genes of the AA metabolism pathway with somatic loss-of-function mutations among 15 OSCC-GB patients and other relevant details.

In addition, genome-wide copy number alteration analysis revealed that there was no statistically significant difference (Fisher’s exact P value=0.623) in the proportion of patients with CN loss in genes of the AAM pathway among those who survived for >12 months (14%) compared with those who did not survive for this length of time (8%; Supplementary Table 4).

Mutational profiles of AAM pathway in oral cancer

The AAM pathway is partitioned into four major sub-components, (I) AA production from membrane phosopholipids by phospholipases and generation of various tumour-inducing inflammatory molecules, like prostaglandins, through metabolism of AA by the action of various (II) COX, leukotrienes by the action of (III) LOX enzymes, and production of several hydroxyeicosatetraenoic acids (HETEs) and epoxyeicosatrienoic acids by the action of (IV) P450 epoxygenase enzymes32. Fifteen non-synonymous somatic single-nucleotide substitutions were identified in 10 genes of the pathway; of which 13 (86%) were predicted to be missense and 2 (14%) were nonsense mutations (Table 1). No somatic insertions or deletions were detected in these genes. These somatic mutations were restricted to specific sub-components of the AAM pathway, which were (a) AA production (four mutated genes—PLA2G3, PLA2G4E, PLA2G4F, PLA2G6), (b) prostaglandin production (mutations in prostacyclin synthase (PTGIS) and TBXAS1), (c) leukotriene production (one mutated gene, GGT7), (d) 5, 19 and 20-HETE production (two mutated genes, CYP2C19 and CYP2U1; see Fig. 1). Two or more samples were discovered to harbour somatic mutations in four genes—GGT7, PTGIS, CYP2U1 and GPX7. GGT7 that are involved in the production of potent cancer inducer LTD4 leukotriene1,33. The enzyme PTGIS catalyses the conversion of prostaglandin H2 (PGH2) to prostaglandin I2 (PGI2). PGI2 promotes cancer growth by activating peroxisome proliferator-activated receptor δ and increases the expression levels of the pro-angiogenic factor vascular endothelial growth factor34,35. CYP2U1 is a hydroxylase that metabolizes AA and produces 20-HETE synthesis, which is associated with the growth of cancer cells in vitro and in vivo and is also upregulated in various human cancers36. Other mutated genes of this pathway are known to aid tumour growth. Of these, GGT7 and PLA2G3 harbour nonsense mutations.

Figure 1: Somatically mutated genes in AAM pathway among patients with OSCC-GB.
figure1

Mutations in AAM pathway genes were mapped on the KEGG pathway map. Genes in the red boxes are those with somatic mutations and genes in green boxes are without any somatic mutation.

Patients with AAM pathway mutations survive longer

The median duration of DFS without recurrence was 14 months (details on Fig. 2). The proportions of patients at advanced stages (T3 or T4) of tumour, lymph node involvement, smoking and chewing habits, alcohol intake, perineural invasion, and bone and skin involvement were not significantly different (P>0.05) between the non-mutated AA (norAA) and mutated AA (mutAA) subgroups of patients. During the follow-up period, a significantly higher proportion (P=0.003) of patients (35 patients; 50.7%) in the norAA subgroup died of cancer than in the mutAA subgroup (one patient; 6%). Mutation in any gene of the AA metabolism was associated with a significant (P=0.003) increase in the median duration of DFS; median durations of DFS in mutAA patients was 24 months, while in norAA patients was 13 months. The Kaplan–Meier survival probability distributions are given in (Fig. 3a).

Figure 2: Clinical, pathological and molecular features of patients without (norAA) and with (mutAA) mutations in genes of the arachidonic acid metabolism (AAM) pathway.
figure2

In the vertical panel, for each patient, the tumor stage (pT), local lymph node metastasis status (pN), HPV infection status, vital status, total number of somatic mutations and mutation status in known OSCC driver genes are shown. These data have been arranged by groups of patients with or without mutations in genes of the AAM pathway in increasing order of the total number of mutations observed in each patient. Colour coding schemes for vital status and driver gene mutation status are indicated on top of the figure.

Figure 3: Kaplan–Meier probability distributions of disease-free survival.
figure3

(a) Subgroups of OSCC-GB patients, with (mutAA) or without (norAA) loss-of-function mutations in genes of the arachidonic acid metabolism pathway; and (b) OSCC-GB patients included in this study pooled with those included in The Cancer Genome Atlas (TCGA) study.

This significant increase in DFS in the mutAA subgroup is not because of the effect of plausible covariates. Recent studies have revealed that (i) intratumour heterogeneity was higher in poor-outcome classes of HNSCC37, and (ii) HPV infection had a favourable outcome among HNSCC patients38,39. HPV infection rates were similar (P=0.72; 21.7% in norAA, 13% in mutAA), the mean number of somatic mutations were similar (P=0.66, t-test of equality of means 95.8±12.5 in mutAA, 89.9±5.6 in norAA. (Two outlier patients with large numbers of somatic mutations, 307 and 622, were excluded from these calculations.) These results were also validated by a multiple regression analysis with DFS as the dependent variable and with independent variables; (a) total number of somatic mutations, (b) whether any gene of AAM pathway was mutated and (c) whether there was HPV infection. The P values corresponding to the three independent variables were, respectively, 1.0, 0.004 and 0.849. Thus, DFS is associated only with mutation in any gene, but not with the total number of mutations or HPV infection status.

Compared with the presence of a AAM pathway mutation, the absence of any mutation on this pathway significantly (P=0.018) increased the hazard ratio for death (11.3; 95% confidence interval 1.5–84.5) even when considered jointly with other variables (age, sex, tumour stage, lymph node involvement and total number of somatic mutations, in a Cox’s proportion hazards model (Supplementary Table 5).

Driver gene profiles are similar in mutAA versus norAA subgroups

The mutAA subgroup had similar mutation profiles in HNSCC/OSCC-GB driver genes—TP53, CASP8, FAT1, NOTCH1, USP9X, ARID2, HRAS, MLL4 and PIK3CA—as the norAA subgroup (Fig. 2). The total number of mutations per patient in these genes was not significantly different (P=0.36) between the norAA (2.39±0.19) and mutAA subgroups (2.0±0.33). Every patient in the mutAA subgroup had at least one somatic mutation in an important OSCC driver gene. The proportions of patients with a somatic mutation in OSCC-GB driver genes were similar in both the subgroups (Supplementary Table 6), except ARID2 and USP9X, which were mutated only in the norAA subgroup (13%). The percentage of patients with somatically mutated MLL4 in the mutAA subgroup (26%) was significantly higher (P=0.01; test of proportions) than that in the norAA subgroup (6%), but the mean number of somatic mutations were similar (P=0.58) in both the subgroups (4.86 in mutAA and 4.31 in norAA). The average mutation rates in DNA repair genes40ATM, ATR, BRCA1, BRCA2, CHEK1, CHEK2, FANCF, MLH1, MSH2, MDC1, PARP1 and RAD51—were similar (P>0.05) in both the subgroups. Further, the proportions of patients with mutations in genes in each mitogenic drug targetable pathway as defined in the KEGG database, namely PI3K–AKT, MAPK, JAK–STAT and WNT signalling, were similar in both the subgroups (P values for all four tests of equality of proportions were >0.05; Supplementary Fig. 1). These results show that the survival advantage enjoyed by patients with somatic mutations in genes of the AAM pathway is not due to confounding factors.

Pooling of TCGA data strengthens inferences

Data of patients with cancer of oral cavity and buccal mucosa (n=50) in the TCGA-HNSCC data set ( https://www.synapse.org/#!Synapse:syn1446094) were mined for the presence of somatic mutations in AAM pathway genes. In the TCGA-OSCC subset, 21 patients (42.0%) had one or more loss-of-function somatic mutations in genes of the AAM pathway (Supplementary Table 7). In the pooled set of OSCC-GB patients included in the present (n=84) and the TCGA studies (n=50), the presence of a mutation in any gene of the AAM pathway was significantly associated with increased DFS (log-rank test; P=0.005). The Kaplan–Meier survival probability distributions are given in Fig. 3b. These inferences also hold if the data on all HNSCC patients included in the TCGA study (n=224) are pooled with the data on OSCC-GB patients (n=84) of the present study. In the TCGA-HNSCC subset, 59 patients (26.0%) had one or more loss-of-function somatic mutations in genes of the AAM pathway (Supplementary Data 1). The total number of mutations was not significantly associated with DFS in the pooled data set either for OSCC-GB (P=0.543) or HNSCC (P=0.905). (Information on HPV infection was not available in the TCGA data set.)

In the pooled (n=308) data set, patients with a mutation in any AAM pathway gene (n=74) had significantly elevated duration of median DFS (39 months) compared with those (n=234) without mutations (29 months; log-rank test; P=0.007). The Kaplan–Meier survival probability distributions for pooled data are given in Supplementary Fig. 2.

PI3K–Akt cascade is downregulated in mutAA patients

Functional proteomics data were also available, in the RRPA data set of the TCGA study, on the HNSCC patients of TCGA considered in the present study. Protein levels between patients with (n=57) or without (n=167) mutated AAM pathway genes were analysed. Differences of protein levels of five genes were found to be statistically significant (P<0.05) between these two groups. Out of these five genes, four genes are known oncogenes (AXL, FN1, MACC1 and IRS1)41,42,43,44,45, which produce significantly low levels of the proteins in patients with mutation in AAM pathway (Supplementary Fig. 3a). Three (AXL, FN1 and IRS1) of these four genes are on the KEGG PI3K–Akt pathway45,46. Further, among all transcription factors, STAT3 and NFKB1 were found to have the lowest levels of phosphorylation among patients with mutations in AAM pathway genes; the levels among patients without mutations were significantly (P<0.05) higher (Supplementary Fig. 3b). Activation of these two transcription factors are known to be mediated by the PI3K–Akt pathway47,48. The PI3K–Akt pathway is known to be activated by AA metabolism pathway end-products6,49 in cancer. PI3K–Akt pathway is a well-known driver of HNSCC50,51.

Compared with the patients without mutation in AAM pathway genes, those with mutation have low protein levels of oncogenes associated with the PI3K–Akt pathway. The low levels of phosphorylation of transcription factors STAT3 and NFKB1, activations of which are mediated by this pathway, also provides indirect support that PI3K–Akt pathway is downregulated among patients with mutated AAM pathway. These are likely explanations of the higher survival benefits observed in HNSCC patients with mutated AAM pathway.

Discussion

Genes on the AAM pathway are known to overexpress and activate other pathways that promote cancer1,52. Targets on this pathway have already been intensively studied in other diseases, such as cardiovascular diseases and pain, providing an opportunity for repurposing of drugs already in clinical development to new indications53.

AA is the central transducer molecule generated from membrane phospholipids by phospholipase enzymes and converted to biologically active tumorigenic eicosanoids by COX, LOX and CYP enzymes. Various genes in this pathway were found to be mutated in a subgroup of patients in this study. The study among Indian OSCC-GB patients has revealed that (a) overall rates and profiles of somatic mutations are similar in the two subgroups of patients with (mutAA) and without (norAA) mutations in genes of the AAM pathway, and (b) the mutAA subgroup of patients harbours mutations in components of AA production and those that induce tumour progression via COX, LOX and CYP enzymes. We have identified missense mutations in four PLA2s family of genes, belonging to groups III, IV and VI, which are involved in the production of AA from membrane phospholipids. Most human PLA2 enzymes overexpress in a proportion of tumours derived from diverse organs including HNSCC54,55,56 and inhibition of PLA2 alters cancer cell growth and death in vitro57. Combination therapy of PLA2 group VI inhibitor and paclitaxel for xenograft model of ovarian cancer have shown promising results58. We have detected two mutations in PTGIS and one mutation in thromboxane A synthase 1 (TBXAS1) gene. A recent study59 has suggested that PTGIS or TBXAS1 may be better therapeutic targets than the most studied and validated target cyclooxygenase-2 (PTGS2) gene in breast cancer. Prostaglandin-targeted therapies (for example, COX inhibitors) are presently under investigation for chemoprevention of a variety of cancers including lung, colorectal and breast cancers60. We have discovered two genes in the LOX pathway—GGT7 and GPX7—to be mutated. Increased LOX expression levels have been observed in a broad range of cancers, including breast, pancreatic, prostate, lung, urinary bladder, leukaemia and colon cancer53. Inhibitors against members of LOX pathway have chemopreventive effects against various cancers61,62,63. Germline variant genotypes in 12-LOX gene reduce the risk of rectal cancer by 33% (ref. 64). Wild-type COX and LOX pathway genes elicit the most potent inflammatory signalling molecules; therefore, blocking these pathways show additive effect in colon cancer cells65. We detected mutations in CYPC219 gene, which is known to overexpress in hepatocellular carcinoma in Chinese patients66, and also in CYP2U1 gene, which is overexpressed in colorectal and ovarian cancers67,68. Combined target inhibition of various combinations of COX, LOX, CYP enzymes as well as inhibition of AA metabolites are effective in human oesophageal, colon, breast and skin cancer1.

Among the mutAA subgroup of patients, the median period of DFS is nearly double of that for the norAA subgroup of patients, which is significant (log-rank test; P=0.003). Further, only 6% patients of the mutAA subgroup died of cancer in contrast to 50.7% of norAA subgroup. These differences were not due to an increase in overall mutation frequency or frequency of HPV infection in the mutAA subgroup, nor even due to statistically significant differences with norAA subgroup in profiles of tumour grade, stage and lymph node involvement or of mutation frequencies in the major genes that drive OSCC-GB14.

Our data combined with HNSCC data of the TCGA study showed that these are globally valid observations. Functional proteomics data revealed that indeed patients with AAM pathway mutations have significantly low protein levels of known oncogenes and low phosphorylation status of nuclear transcription factors, which play critical roles in the PI3K–Akt cascade. PI3k–Akt pathway is a known driver of HNSCC. Downregulation of this cascade in patients with mutated AAM pathway is congruent with their observed higher DFS and are likely to be the explanatory factors.

Our results provide the definitive inference that presence of mutations in genes of AA metabolism pathway may be used as a prognostic marker of survival after treatment of GB oral cancer.

Methods

Ethics approval and informed consent

Institutional Ethics Committees of the Advanced Centre for Treatment, Research and Education in Cancer (ACTREC) and the National Institute of Biomedical Genomics (NIBMG) approved this study. Voluntary informed consent was obtained from each participant.

Patient description

Blood and tumour specimens were collected at surgical excision, from 110 treatment-naïve OSCC-GB patients, who underwent a comprehensive staging, curative resection, post-operative radiotherapy±chemotherapy and follow-up at ACTREC. Only patients with concordant histologic diagnoses by two independent and experienced histopathologists, who followed the WHO guidelines, were recruited into this study. Each patient was also followed up every month after surgery. At each follow-up, clinical checks of the primary site and relevant nodes were carried out for recurrence, and a verbal reporting of the health status was obtained from the patient. Clinical detection of recurrence was confirmed by histopathology. We have reported only cancer-specific deaths. Patients who died of causes other than cancer were excluded from the analyses. DFS time was estimated as the maximum follow-up period during which a patient remained free of OSCC after treatment. Subsequently, OSCC may have recurred in the patient or the patient may have died of other causes.

DNA isolation, exome sequencing, genotyping and HPV assay

DNA isolation, determination of HPV infection status—exome sequencing using three orthogonal technology platforms and whole-genome Illumina SNParray genotyping were done as described previously14. In brief, for exome sequencing, about 62 Mb of coding region of the human genome was captured using Illumina Tru-Seq exome enrichment kit and sequenced using Illumina Hiseq 2000. DNA samples were also captured for ~26 Mb of coding region by using Nimblegen SeqCap EZ and sequenced using Roche GS-FLX platforms to verify variants identified from the Illumina study. In addition, mutated variants from outside of the Nimblegen ~26 Mb capture region were verified using the Ion Torrent platform. Each blood and tumour exome library was sequenced at a minimum depth of 25 × and 35 × , respectively.

Somatic mutation and copy number alteration determination, and verification

Sequence data obtained using HiSeq-2000 were demultiplexed and FASTQ files were generated by CASAVA v1.8.2 (Illumina). FASTQC Ver. 0.10.1 ( http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) was used to quality-check sequence data, using default parameters. Sequence reads with >10% N-bases were filtered and removed. Sequence reads with average quality value <20 were also removed. Filtered sequence reads were mapped to human genome reference sequence (hg19 with decoy sequence; ftp://ftp.1000genomes.ebi.ac.uk/vol1/ftp/technical/reference/phase2_reference_ assembly_sequence/) using BWA v0.6.1 with –q 15, −l 32 as arguments and SAM files were generated. Reads that mapped to decoy segments of the reference sequence and those that multiple mapped to various locations were removed from further analysis. Finally, unmapped and duplicate reads were filtered out using programs written by us and PICARD v1.60 (http://picard.sourceforge.net). GATK v. 1.4. was used for base quality recalibration, local realignment around indels using known single-nucleotide polymorphisms. Alignment SAM files were converted to BAM files using Samtools 0.1.12a. BAM files were converted to pileup format by SAMtools (http://samtools.sourceforge.net/) for variant calling. Somatic variants were called using a base-by-base caller14. High confidence variants were identified using the following criteria, (i) variant base quality value >15, (ii) minimum depth ≥10 in blood and tumour, (iii) for a site to be scored as a loss of heterozygosity site, at least 20% of the reads that map to the site should contain the variant in the blood DNA; (iv) at least 10% of the reads should contain a variant in the tumour DNA for a site to be scored as a somatic variant site, (v) a variant site should have at least one read with the variant in each direction (to avoid strand bias), (vi) for every patient, clustered somatic mutations (two mutations in a 100-bp window) were removed, as these mutations were possibly generated as part of read misalignment error and (viii) somatic variants which are catalogued in the 1,000 Genomes, ESP genome database (https://esp.gs.washington.edu/) with an allele frequency >0.01 were removed. If at a site both somatic and germline mutations were detected in our cohort, then the site was removed. Somatic mutations thus identified were annotated using the Annovar package and Oncotator module (http://www.broadinstitute.org/ cancer/cga/oncotator) to infer the locations of the variants within the genes, effect of the mutation on the protein product and other features. All indels and recurrent substitutions were manually visualized and verified using IGV v2.1.25.

Sequence data generated using GS-FLX Genome Sequencers, were converted to FASTQ format from SFF files by sffinfo (GS-FLX) and a Linux script developed by us. For data generated by Ion-torrent PGM, sequence reads were obtained in FASTQ format. The sequence data generated per run were evaluated by the same method described in the previous section. The GS-FLX derived FASTQ files were mapped to the human reference sequence hg19 (GRCh37) by SSAHA2 v2.5.5. PICARD v1.60 (http://picard.sourceforge.net) was used to remove PCR duplicates and combine SAM files from multiple runs. Pileup files were generated by SAMtools package. The pileup files of tumour and blood DNA from each patient were used to identify and verify putative variant sites using the base-by-base variant caller. For CNV, ASCAT 2.0 (alelle-specific copy number analysis of tumours) was used to identify regions of amplification and deletion using exported data from Illumina iScan Genome Studio software. Segments with mean LogR>0.15 were used to detect amplification segment and mean LogR<−0.2 were used to detect deletion. High-confidence CNV regions were annotated with NCBI refgene transcript database (hg19).

Sequence BAM files for patients have been deposited in the European Genome-Phenome Archive under accession codes EGAS00001000249 and EGAS00001001028.

Pathway analysis and other bioinformatics analyses

PathScan module in GenomeMuSiC package69 was used to identify somatic mutations enriched in annotated KEGG (Kyoto Encyclopedia of Genes and Genomes) AA metabolism pathway and other important mitogenic signalling pathways (PI3K, MAPK, JAK-STAT, WNT signalling). Only the loss-of-function mutations (non-synonymous, frame-shift and splice-site) were used in the PathScan analysis. Patients possessing a non-synonymous somatic mutation in a gene that overlapped with the Cancer Gene Census list (COSMIC Database) were identified. Similarly, the list of druggable genes70 in mitogenic signalling pathways was searched for identifying the number of cancer genes mutated. Clinical and survival information for HNSCC and OSCC-GB patients included in the TCGA project were retrieved from the Synapse website https://www.synapse.org/#!Synapse:syn1446094. Somatic mutation data of the corresponding patients for the AA metabolism pathway genes were downloaded from http://www.cbioportal.org/public-portal/ website.

Analysis of TCGA-HNSCC functional proteomics data

Custom search was done with 224 HNSCC patient IDs and AA metabolism pathway gene list on the cbioportal for cancer genomics (http://www.cbioportal.org/public-portal/) for comparative analyses of mutation and reverse phase protein array data. Reverse phase protein array is a rapid high-throughput antibody-based technique to detect cellular protein levels; the procedure is very similar to western blotting. The custom analysis framework in cBioPortal produces a list of the significantly altered proteins based on their levels between patients with mutations in the list of mutated genes versus those with no mutations in genes on this list. A two-sided two sample Student’s t-test is used to detect significant protein alteration.

Statistical tests

Estimates of period of DFS for patients with mutAA genes versus patients with norAA genes were compared using the log-rank test. The survival probability distributions for both the groups were estimated using the Kaplan–Meier method. Two-sided statistical tests were used. Multivariate Cox proportional-hazards modelling was used to assess the possible prognostic significance of AAM pathway mutation on DFS. As all the patients were exposed to tobacco, tobacco use was not included as a covariate. In the Cox proportional-hazards analysis, age and number of somatic mutations were included as continuous variables, whereas all other factors were considered as categorical. P values for hazard ratios were calculated using the likelihood ratio test. P values <0.05 were considered to be significant.

Additional information

How to cite this article: Biswas, N. K. et al. Somatic mutations in arachidonic acid metabolism pathway genes enhance oral cancer post-treatment disease-free survival. Nat. Commun. 5:5835 doi: 10.1038/ncomms6835 (2014).

Accession codes: Sequence BAM files for patients have been deposited in the European Genome-Phenome Archive under accession code EGAS00001001028.

References

  1. 1

    Wang, D. Z. & Dubois, R. N. Eicosanoids and cancer. Nat. Rev. Cancer 10, 181–193 (2010).

    CAS  Article  Google Scholar 

  2. 2

    Gorrini, C., Harris, I. S. & Mak, T. W. Modulation of oxidative stress as an anticancer strategy. Nat. Rev. Drug Discov. 12, 931–947 (2013).

    CAS  Article  Google Scholar 

  3. 3

    Marnett, L. J. Generation of mutagens during arachidonic acid metabolism. Cancer Metastasis Rev. 13, 303–308 (1994).

    CAS  Article  Google Scholar 

  4. 4

    Leslie, C. C. Regulation of arachidonic acid availability for eicosanoid production. Biochem. Cell Biol. 82, 1–17 (2004).

    CAS  Article  Google Scholar 

  5. 5

    Piomelli, D. Arachidonic acid in cell signalling. Curr. Opin. Cell Biol. 5, 274–280 (1993).

    CAS  Article  Google Scholar 

  6. 6

    Hughes-Fulford, M., Li, C. F., Boonyaratanakornkit, J. & Sayyah, S. Arachidonic acid activates phosphatidylinositol 3-kinase signaling and induces gene expression in prostate cancer. Cancer Res. 66, 1427–1433 (2006).

    CAS  Article  Google Scholar 

  7. 7

    Paine, E., Palmantier, R., Akiyama, S. K., Olden, K. & Roberts, J. D. Arachidonic acid activates mitogen-activated protein (MAP) kinase-activated protein kinase 2 and mediates adhesion of a human breast carcinoma cell line to collagen type IV through a p38 MAP kinase-dependent pathway. J. Biol. Chem. 275, 11284–11290 (2000).

    CAS  Article  Google Scholar 

  8. 8

    Cui, X. L. & Douglas, J. G. Arachidonic acid activates c-jun N-terminal kinase through NADPH oxidase in rabbit proximal tubular epithelial cells. Proc. Natl Acad. Sci. USA 94, 3771–3776 (1997).

    CAS  ADS  Article  Google Scholar 

  9. 9

    Ondrey, F. G. Arachidonic acid metabolism: a primer for head and neck surgeons. Head Neck 20, 334–349 (1998).

    CAS  Article  Google Scholar 

  10. 10

    Karmali, R. A., Wustrow, T., Thaler, H. T. & Strong, E. W. Prostaglandins in carcinomas of the head and neck. Cancer Lett. 22, 333–336 (1984).

    CAS  Article  Google Scholar 

  11. 11

    Jung, T. T., Berlinger, N. T. & Juhn, S. K. Prostaglandins in squamous cell carcinoma of the head and neck: a preliminary study. Laryngoscope 95, 307–312 (1985).

    CAS  Article  Google Scholar 

  12. 12

    Koontongkaew, S., Monthanapisut, P. & Saensuk, T. Inhibition of arachidonic acid metabolism decreases tumor cell invasion and matrix metalloproteinase expression. Prostaglandins Other Lipid Mediat. 93, 100–108 (2010).

    CAS  Article  Google Scholar 

  13. 13

    El-Hakim, I. E. & Langdon, J. D. Arachidonic acid cascade and oral squamous cell carcinoma. Clin. Otolaryngol. Allied Sci. 16, 563–573 (1991).

    CAS  Article  Google Scholar 

  14. 14

    India Project Team of the International Cancer Genome Consortium. Mutational landscape of gingivo-buccal oral squamous cell carcinoma reveals new recurrently-mutated genes and molecular subgroups. Nat. Commun. 4, 2873 (2013).

  15. 15

    Warnakulasuriya, S. Global epidemiology of oral and oropharyngeal cancer. Oral Oncol. 45, 309–316 (2009).

    Article  Google Scholar 

  16. 16

    Khan, Z. An overview of oral cancer in Indian subcontinent and recommendations to decrease its incidence. WebmedCentral Cancer 3, WMC003626 (2012) ) http://www.webmedcentral.com/article_view/3626 (accessed August 2013).

    Google Scholar 

  17. 17

    Walvekar, R. R. Squamous cell carcinoma of the gingivobuccal complex: predictors of locoregional failure in stage III-IV cancers. Oral Oncol. 45, 135–140 (2009).

    CAS  Article  Google Scholar 

  18. 18

    Lippman, S. M. & Hong, W. K. Molecular markers of the risk of oral cancer. New Engl. J. Med. 344, 1323–1326 (2001).

    CAS  Article  Google Scholar 

  19. 19

    Agrawal, N. et al. Exome sequencing of head and neck squamous cell carcinoma reveals inactivating mutations in NOTCH1. Science 333, 1154–1157 (2011).

    CAS  ADS  Article  Google Scholar 

  20. 20

    Stransky, N. et al. The mutational landscape of head and neck squamous cell carcinoma. Science 333, 1157–1160 (2011).

    CAS  ADS  Article  Google Scholar 

  21. 21

    Pickering, C. R. et al. Integrative genomic characterization of oral squamous cell carcinoma identifies frequent somatic drivers. Cancer Discov. 3, 770–781 (2013).

    CAS  Article  Google Scholar 

  22. 22

    Fanjul-Fernández, M. et al. Cell–cell adhesion genes CTNNA2 and CTNNA3 are tumour suppressors frequently mutated in laryngeal carcinoma. Nat. Commun. 4, 2531 (2013).

    Article  Google Scholar 

  23. 23

    Sveinbjörnsson, B. et al. Expression of enzymes and receptors of the leukotriene pathway in human neuroblastoma promotes tumor survival and provides a target for therapy. FASEB J. 22, 3525–3536 (2008).

    Article  Google Scholar 

  24. 24

    Zhi, H. et al. The deregulation of arachidonic acid metabolism-related genes in human esophageal squamous cell carcinoma. Int. J. Cancer 106, 327–333 (2003).

    CAS  Article  Google Scholar 

  25. 25

    Möbius, C. et al. Prognostic value of eicosanoid pathways in extrahepatic cholangiocarcinoma. Anticancer Res. 28, 873–878 (2008).

    PubMed  Google Scholar 

  26. 26

    Panigrahy, D., Kaipainen, A., Greene, E. R. & Huang, S. Cytochrome P450-derived eicosanoids: the neglected pathway in cancer. Cancer Metastasis Rev. 29, 723–735 (2010).

    CAS  Article  Google Scholar 

  27. 27

    Ghosh, J. & Myers, C. E. Inhibition of arachidonate 5-lipoxygenase triggers massive apoptosis in human prostate cancer cells. Proc. Natl Acad. Sci. USA 95, 13182–13187 (1998).

    CAS  ADS  Article  Google Scholar 

  28. 28

    Kim, J. H., Hubbard, N. E., Ziboh, V. & Erickson, K. L. Attenuation of breast tumor cell growth by conjugated linoleic acid via inhibition of 5-lipoxygenase activating protein. Biochim. Biophys. Acta 1736, 244–250 (2005).

    CAS  Article  Google Scholar 

  29. 29

    Park, S. W., Heo, D. S. & Sung, M. W. The shunting of arachidonic acid metabolism to 5-lipoxygenase and cytochrome p450 epoxygenase antagonizes the anti-cancer effect of cyclooxygenase-2 inhibition in head and neck cancer cells. Cell Oncol. (Dordr) 35, 1–8 (2012).

    Article  Google Scholar 

  30. 30

    Saba, N. F. et al. Chemoprevention of head and neck cancer with celecoxib and erlotinib: results of a phase ib and pharmacokinetic study. Cancer Prev. Res. (Phila) 7, 283–291 (2014).

    CAS  Article  Google Scholar 

  31. 31

    Zagorac, D., Jakovcevic, D., Gebremedhin, D. & Harder, D. R. Antiangiogenic effect of inhibitors of cytochrome P450 on rats with glioblastoma multiforme. J. Cereb. Blood Flow Metab. 28, 1431–1439 (2008).

    CAS  Article  Google Scholar 

  32. 32

    Needleman, P., Turk, J., Jakschik, B. A., Morrison, A. R. & Lefkowith, J. B. Arachidonic acid metabolism. Annu. Rev. Biochem. 55, 69–102 (1986).

    CAS  Article  Google Scholar 

  33. 33

    Heisterkamp, N., Groffen, J., Warburton, D. & Sneddon, T. P. The human gamma-glutamyltransferase gene family. Hum. Genet. 4, 321–332 (2008).

    Article  Google Scholar 

  34. 34

    Wang, J. et al. VEGF expression is augmented by hypoxia induced PTGIS in human fibroblasts. Int. J. Oncol. 43, 746–754 (2013).

    CAS  Article  Google Scholar 

  35. 35

    Lichao, S. et al. Overexpression of PTGIS could predict liver metastasis and is correlated with poor prognosis in colon cancer patients. Pathol. Oncol. Res. 18, 563–569 (2012).

    Article  Google Scholar 

  36. 36

    Alexanian, A., Miller, B., Roman, R. J. & Sorokin, A. 20-HETE-producing enzymes are up-regulated in human cancers. Cancer Genomics Proteomics 9, 163–169 (2012).

    CAS  PubMed  PubMed Central  Google Scholar 

  37. 37

    Mroz, E. A. & Rocco, J. W. MATH, a novel measure of intratumor genetic heterogeneity, is high in poor-outcome classes of head and neck squamous cell carcinoma. Oral Oncol. 3, 211–215 (2013).

    Article  Google Scholar 

  38. 38

    Jo, S. et al. Human papillomavirus infection as a prognostic factor in oropharyngeal squamous cell carcinomas treated in a prospective phase II clinical trial. Anticancer Res. 5, 1467–1474 (2009).

    Google Scholar 

  39. 39

    Nagel, R. et al. Treatment response of HPV-positive and HPV-negative head and neck squamous cell carcinoma cell lines. Oral Oncol. 6, 560–566 (2013).

    Article  Google Scholar 

  40. 40

    Cerami, E. et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov. 2, 401–404 (2012).

    Article  Google Scholar 

  41. 41

    Paccez, J. D., Vogelsang, M., Parker, M. I. & Zerbini, L. F. The receptor tyrosine kinase Axl in cancer: biological functions and therapeutic implications. Int. J. Cancer 134, 1024–1033 (2014).

    CAS  Article  Google Scholar 

  42. 42

    Lee, C. H. et al. Axl is a prognostic marker in oral squamous cell carcinoma. Ann. Surg. Oncol. 3, 500–508 (2012).

    Article  Google Scholar 

  43. 43

    Hynes, R. O., Destree, A. T., Perkins, M. E. & Wagner, D. D. Cell surface fibronectin and oncogenic transformation. J. Supramol. Struct. 11, 95–104 (1979).

    CAS  Article  Google Scholar 

  44. 44

    Huang, Y. et al. Overexpression of MACC1 and Its significance in human Breast Cancer Progression. Cell Biosci. 3, 16 (2013).

    CAS  Article  Google Scholar 

  45. 45

    Hao, Y. et al. Gain of interaction with IRS1 by p110α-helical domain mutants is crucial for their oncogenic functions. Cancer Cell. 23, 583–593 (2013).

    CAS  Article  Google Scholar 

  46. 46

    Braunger, J. et al. Intracellular signaling of the Ufo/Axl receptor tyrosine kinase is mediated mainly by a multi-substrate docking-site. Oncogene 14, 2619–2631 (1997).

    CAS  Article  Google Scholar 

  47. 47

    Bai, D., Ueno, L. & Vogt, P. K. Akt-mediated regulation of NFkappaB and the essentialness of NFkappaB for the oncogenicity of PI3K and Akt. Int. J. Cancer 125, 2863–2870 (2009).

    CAS  Article  Google Scholar 

  48. 48

    Hart, J. R., Liao, L., Yates, J. R. 3rd. & Vogt, P. K. Essential role of Stat3 in PI3K-induced oncogenic transformation. Proc. Natl Acad. Sci. USA 108, 13247–13252 (2011).

    CAS  ADS  Article  Google Scholar 

  49. 49

    Hii, C. S., Moghadammi, N., Dunbar, A. & Ferrante, A. Activation of the phosphatidylinositol 3-kinase-Akt/protein kinase B signaling pathway in arachidonic acid-stimulated human myeloid and endothelial cells: involvement of the ErbB receptor family. J. Biol. Chem. 276, 27246–27255 (2001).

    CAS  Article  Google Scholar 

  50. 50

    Broek, R. V., Mohan, S., Eytan, D. F., Chen, Z. & Van, W. C. The PI3K/Akt/mTOR axis in head and neck cancer: functions, aberrations, crosstalk, and therapies. The PI3K/Akt/mTOR axis in head and neck cancer: functions, aberrations, crosstalk, and therapies. Oral Dis. doi:10.1111/odi.12206 (2013).

  51. 51

    Iglesias-Bartolome, R., Martin, D. & Gutkind, J. S. Exploiting the head and neck cancer oncogenome: widespread PI3K-mTOR pathway alterations and novel molecular targets. Cancer Discov. 3, 722–725 (2013).

    CAS  Article  Google Scholar 

  52. 52

    Marks, F., Fürstenberger, G. & Müller-Decker, K. Tumor promotion as a target of cancer prevention. Recent Results Cancer Res. 174, 37–47 (2007).

    CAS  Article  Google Scholar 

  53. 53

    Hyde, C. A. & Missailidis, S. Inhibition of arachidonic acid metabolism and its implication on cell proliferation and tumour-angiogenesis. Int. Immunopharmacol. 9, 701–715 (2009).

    CAS  Article  Google Scholar 

  54. 54

    Linkous, A. G., Yazlovitskaya, E. M. & Hallahan, D. E. Cytosolic phospholipase A2 and lysophospholipids in tumor angiogenesis. J. Natl Cancer Inst. 102, 1398–1412 (2010).

    CAS  Article  Google Scholar 

  55. 55

    Tripathi, P. et al. Delineating metabolic signatures of head and neck squamous cell carcinoma: phospholipase A2, a potential therapeutic target. Int. J. Biochem. Cell Biol. 44, 1852–1861 (2012).

    CAS  Article  Google Scholar 

  56. 56

    Scott, K. F. et al. Emerging roles for phospholipase A2 enzymes in cancer. Biochimie. 92, 601–610 (2010).

    CAS  Article  Google Scholar 

  57. 57

    Cummings, B. S. Phospholipase A2 as targets for anti-cancer drugs. Biochem. Pharmacol. 74, 949–959 (2007).

    CAS  Article  Google Scholar 

  58. 58

    Li, H. et al. Combination therapy of an inhibitor of group VIA phospholipase A2 with paclitaxel is highly effective in blocking ovarian cancer development. Am. J. Pathol. 179, 452–461 (2011).

    CAS  Article  Google Scholar 

  59. 59

    Abraham, J. E. et al. Common polymorphisms in the prostaglandin pathway genes and their association with breast cancer susceptibility and survival. Clin. Cancer Res. 15, 2181–2191 (2009).

    CAS  Article  Google Scholar 

  60. 60

    DuBois, R. N. Evaluation of the whole prostaglandin biosynthetic pathway in lung cancer. Clin Cancer Res. 9, 1577–1578 (2003).

    PubMed  Google Scholar 

  61. 61

    Chen, X. et al. Overexpression of 5-lipoxygenase in rat and human esophageal adenocarcinoma and inhibitory effects of zileuton and celecoxib on carcinogenesis. Clin. Cancer Res. 10, 6703–6709 (2004).

    CAS  Article  Google Scholar 

  62. 62

    Ding, X. Z., Kuszynski, C. A., El-Metwally, T. H. & Adrian, T. E. Lipoxygenase inhibition induced apoptosis, morphological changes, and carbonic anhydrase expression in human pancreatic cancer cells. Biochem. Biophys. Res. Commun. 266, 392–399 (1999).

    CAS  Article  Google Scholar 

  63. 63

    Wong, B. C. et al. 12-Lipoxygenase inhibition induced apoptosis in human gastric cancer cells. Carcinogenesis 22, 1349–1354 (2001).

    CAS  Article  Google Scholar 

  64. 64

    Li, S. et al. Polymorphisms in arachidonic acid metabolism-related genes and the risk and prognosis of colorectal cancer. Fam. Cancer. 12, 755–765 (2013).

    CAS  Article  Google Scholar 

  65. 65

    Cianchi, F. et al. Inhibition of 5-lipoxygenase by MK886 augments the antitumor activity of celecoxib in human colon cancer cells. Mol. Cancer Ther. 5, 2716–2726 (2006).

    CAS  Article  Google Scholar 

  66. 66

    Wu, M., Chen, S. & Wu, X. Differences in cytochrome P450 2C19 (CYP2C19) expression in adjacent normal and tumor tissues in Chinese cancer patients. Med. Sci. Monit. 12, 174–178 (2006).

    Google Scholar 

  67. 67

    Downie, D. et al. Profiling cytochrome P450 expression in ovarian cancer: identification of prognostic markers. Clin. Cancer Res. 11, 7369–7375 (2005).

    CAS  Article  Google Scholar 

  68. 68

    Kumarakulasingham, M. et al. Cytochrome p450 profile of colorectal cancer: identification of markers of prognosis. Clin. Cancer Res. 11, 3758–3765 (2005).

    CAS  Article  Google Scholar 

  69. 69

    Dees, N. D. et al. MuSiC: identifying mutational significance in cancer genomes. Genome Res. 22, 1589–1598 (2012).

    CAS  Article  Google Scholar 

  70. 70

    Lui, V. W. et al. Frequent mutation of the PI3K pathway in head and neck cancer defines predictive biomarkers. Cancer Discov. 3, 761–769 (2013).

    CAS  Article  Google Scholar 

Download references

Acknowledgements

We thank all the members of the Indian project team of the International Cancer Genome Consortium. Special thanks are due to Dr Analabha Basu and Mr Shantanu Kumar for discussions and help at various stages of this work. We are grateful to the Department of Biotechnology, Government of India, for providing financial and logistical support to the Indian project on gingivo-buccal oral cancer as a part of the ICGC.

Author information

Affiliations

Authors

Contributions

N.K.B. conceived the study; N.K.B. and P.P.M. wrote the manuscript; R.S. coordinated patient recruitment, clinical assessments and collection of follow-up data on patients; A.M. guided and coordinated massively parallel sequencing and other experiments; P.P.M. provided overall, including statistical, guidance; and, N.K.B. and S.D. carried out statistical and bioinformatics analysis.

Corresponding author

Correspondence to Nidhan K. Biswas.

Ethics declarations

Competing interests

The authors declare no competing financial interests.

Supplementary information

Supplementary Information

Supplementary Figures 1-3 and Supplementary Tables 1-7. (PDF 1286 kb)

Supplementary Data Set 1

Vital status, surival period and distribution of somatic loss-of-function mutations the HNSCC subset (n=224) of patients included in the TCGA study. (XLS 55 kb)

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Biswas, N., Das, S., Maitra, A. et al. Somatic mutations in arachidonic acid metabolism pathway genes enhance oral cancer post-treatment disease-free survival. Nat Commun 5, 5835 (2014). https://doi.org/10.1038/ncomms6835

Download citation

Further reading

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing