Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Genetic variation in the eicosanoid pathway is associated with non-small-cell lung cancer (NSCLC) survival

  • Lindsay N. Sausville,

    Roles Conceptualization, Formal analysis, Investigation, Writing – original draft

    Affiliations Department of Genetics, Geisel School of Medicine, Dartmouth College, Hanover, NH, United States of America, Department of Epidemiology and Biostatistics, Case Western Reserve University, Cleveland, OH, United States of America

  • Carissa C. Jones,

    Roles Data curation, Formal analysis, Resources, Writing – review & editing

    Affiliations Vanderbilt Genetics Institute, Vanderbilt University Medical Center, Nashville, TN, United States of America, Department of Thoracic Surgery, Vanderbilt University Medical Center, Nashville, TN, United States of America

  • Melinda C. Aldrich,

    Roles Conceptualization, Investigation, Resources, Supervision, Writing – review & editing

    Affiliations Vanderbilt Genetics Institute, Vanderbilt University Medical Center, Nashville, TN, United States of America, Department of Thoracic Surgery, Vanderbilt University Medical Center, Nashville, TN, United States of America, Division of Epidemiology, Vanderbilt University School of Medicine, Nashville, TN, United States of America

  • William J. Blot,

    Roles Conceptualization, Funding acquisition, Investigation, Resources, Supervision, Writing – review & editing

    Affiliations Division of Epidemiology, Vanderbilt University School of Medicine, Nashville, TN, United States of America, International Epidemiology Institute, Rockville, MD, United States of America

  • Ambra Pozzi ,

    Contributed equally to this work with: Ambra Pozzi, Scott M. Williams

    Roles Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing – original draft, Writing – review & editing (SMW); (AP)

    Affiliations Department of Medicine, Vanderbilt University Medical Center, Nashville, TN, United States of America, Department of Medicine, Veterans Affairs Hospital, Nashville, TN, United States of America

  • Scott M. Williams

    Contributed equally to this work with: Ambra Pozzi, Scott M. Williams

    Roles Conceptualization, Formal analysis, Funding acquisition, Project administration, Supervision, Writing – original draft, Writing – review & editing (SMW); (AP)

    Affiliations Department of Genetics, Geisel School of Medicine, Dartmouth College, Hanover, NH, United States of America, Department of Epidemiology and Biostatistics, Case Western Reserve University, Cleveland, OH, United States of America


Genetic variation in the eicosanoid pathway is associated with non-small-cell lung cancer (NSCLC) survival

  • Lindsay N. Sausville, 
  • Carissa C. Jones, 
  • Melinda C. Aldrich, 
  • William J. Blot, 
  • Ambra Pozzi, 
  • Scott M. Williams


Globally, lung cancer results in more deaths worldwide than any other cancer, indicating a need for better treatments. Members of the eicosanoid metabolism pathway represent promising therapeutic targets, as several enzymes involved in the generation of these lipids are dysregulated in many cancers and their inhibition reduces lung cancer growth in mouse models. However, genetic variation of enzymes involved in eicosanoid metabolism has not been adequately examined for association with lung cancer. The goal of this study was to determine whether germline genetic variation altering eicosanoid producing enzyme function and/or expression are associated with differences in lung cancer survival. We examined the association of genetic variation with mortality within eicosanoid metabolism genes in 395 non-small-cell lung cancer (NSCLC) cases from the Southern Community Cohort Study (SCCS). A total of 108 SNPs, both common and rare, in 19 genes, were examined for association. No common or rare variants were associated with lung cancer survival across the entire study population. However, rare variants in ALOX15B (arachidonate 15-lipoxygenase, type B) and the common variant rs12529 in AKR1C3 (prostaglandin F synthase) were associated with NSCLC mortality in women and African Americans, respectively. Rare variants in ALOX15B were associated with greater mortality in women (HR = 2.10, 95% CI = 1.25–3.54, p-value = 0.005). The major allele of rs12529 in AKCR1C3 associated with improved survival in African Americans (HR = 0.74, 95% CI = 0.59–0.92, p-value = 0.008). The lack of genetic associations among all NSCLC cases and the association among women only for rare variants in ALOX15B may, in part, explain the better NSCLC survival observed among women. These results raise the possibility that some subgroups within the NSCLC population may benefit from drugs targeting eicosanoid metabolism.


Lung cancer causes more than one million deaths annually worldwide, with most cases being non-small-cell lung cancer (NSCLC) [13]. A major cause of the high mortality is that most patients present with advanced-stage disease, precluding the possibility of successful surgical resection. In this case alternative treatment modalities, such as chemotherapy and radiation, are used, but they are less effective than resection of localized disease [4, 5]. The late stage of diagnosis and the aggressive nature of NSCLC together lead to a 5-year survival of only 18% and clearly point to the need for more effective therapies [5]. To target NSCLC for treatment it is necessary to better understand processes that affect tumor progression.

Broadly, tumor cells rely on the release of autocrine and paracrine factors to promote tumor growth and to alter tumor microenvironment (TME) responses. In turn, TME cells, such as endothelial cells, fibroblasts, and macrophages can secrete factors affecting tumor cell response, and this interplay regulates essential processes such as cancer cell proliferation, migration, neovascularization, and metastasis [6]. A major group of secreted factors altering these processes are the eicosanoids. These molecules are produced by either the tumor cells or TME cells, thus affecting processes such as inflammation, angiogenesis, and tumor cell growth/dissemination [7].

Eicosanoid metabolites are derived from the metabolism of arachidonic acid through three major branches: cyclooxygenase (COX), lipoxygenase (LOX), and the cytochrome P450 monooxygenase branches (Fig 1A). Each of these branches generates specific eicosanoid metabolites having different biological effects. Genes in these branches that were assayed in our study are denoted in Fig 1B. The COX-derived products begin with the generation of the prostaglandin endoperoxide prostaglandin H2 (PGH2) by COX-1 or COX-2. Selective prostaglandin synthases then convert this intermediate into corresponding prostanoids including thromboxane (TXA2), prostaglandin D2 (PGD2), prostaglandin E2 (PGE2), prostaglandin F2 (PGF), and prostaglandin I2 (PGI2) (Fig 2; S1 Table). The LOX pathway metabolizes arachidonic acid to hydro(pero)xyeicosatetraenoic acids that can undergo further reactions to generate 5-, 8-, 12-, and 15-hydroeicosatetraenoic acids (HETEs), leukotrienes, and lipoxins (Fig 3). Finally, the cytochrome P450 monooxygenases metabolize arachidonic acid to epoxyeicosatrienoic acids (EETs) or 20-HETE (Fig 4) [7].

Fig 1. Eicosanoid metabolism.

(A) Eicosanoid metabolite generation begins with the metabolism of arachidonic acid by cyclooxygenase (COX), lipooxygenase (LOX), or the cytochrome P450 monooxygenases, which includes both the cytochrome P450 epoxygenases and cytochrome p450 ω-hydroxylases. Each of these enzymes produces its own major metabolites displayed here according to the first enzyme acting on arachidonic acid. Abbreviations are as followed: prostaglandin D2 (PGD2), prostaglandin E2 (PGE2), prostaglandin F (PGF), prostaglandin I2 (PGI2), thromboxane (TXA2), hydroxyeicosatetraenoic acids (HETEs), and epoxyeicosatrienoic acids (EETs). (B) The genes encoding the enzymes responsible for metabolism of major metabolites are shown here with the number of common and rare SNPs in blue and gray, respectively.

Fig 2. Metabolism of COX-derived eicosanoids.

Arachidonic acid is first metabolized by COX-1 or COX-2 to synthesize PGH2. Five different types of prostaglandin synthases act on this intermediate to generate the various eicosanoids shown above. In addition, subsequent dehydration/isomerization of PGD2 can create 15d-PGJ2. With the exception of three eicosanoids, 15d-PGJ2, PGD2, and PGI2, all of the COX-derived eicosanoids are thought to be pro-tumorigenic. COX-derived eicosanoids in green are pro-tumorigenic, while those in blue are anti-tumorigenic. Abbreviations are as followed: 15-deoxy-Δ12,14-PGJ2 (15d-PGJ2), arachidonic acid (AA), prostaglandin D2 (PGD2), prostaglandin E2 (PGE2), prostaglandin F (PGF), prostaglandin H2 (PGH2), prostaglandin I2 (PGI2), and thromboxane A2 (TXA2).

Fig 3. Metabolism of LOX-derived eicosanoids.

The initial reaction in the LOX pathway can be carried out by 5-LOX, 12-LOX, or 15-LOX, with each LOX synthesizing its own corresponding HpETE. Subsequent reduction of these HpETEs generates HETEs. However, 5-HpETE and 15-HpETE can undergo other reactions as well. More specially, 5-LOX further metabolizes 5-HpETE to LTA4 that serves a precursor for LXs, LTB4, or cysteinyl leukotrienes. 5-LOX can also act on 15-HpETE to synthesize LXs. LOX-derived eicosanoids are color coded to indicate their effect on tumors. Blue metabolites are anti-tumorigeneic, green metabolites are pro-tumorigenic, and gray metabolites are unknown. Abbreviations are as followed: arachidonic acid (AA), hydroperoxyeicosatetraenoic acid (HpETE), hydroxyeicosatetraenoic acid (HETE), leukotriene A4 (LTA4), leukotriene B4 (LTB4), and lipoxins (LXs).

Fig 4. Metabolism of CYP450 monooxygenase-derived eicosanoids.

Arachidonic acid can enter either of two CYP450 monooxygenase arms. The CYP450 epoxygenase arm generates EETs, while the CYP450 ω-hydroxylation arm generates 20-HETE. Abbreviations are as followed: arachidonic acid (AA), epoxyeicosatrienoic acids (EETs), and hydroxyeicosatetraenoic acid (HETE).

Many of the above mentioned enzymes play a role in tumor biology based on several observations. Expression of many eicosanoid producing genes is frequently dysregulated in tumors. For example, upregulation of COX-2 expression is common in multiple tumors types, including NSCLC, and is associated with poor prognosis by enhancing processes such as angiogenesis [813]. Although COX-2 inhibitors are successful in suppressing the development and progression of lung cancer in animal models, increased cardiovascular risks of COX-2 inhibitors limit their use in chemoprevention of cancer. To obviate these obstacles, blocking COX-2 activity or expression by enhancing levels of glucocorticoids, endogenous and potent COX-2 inhibitors, can suppress colon and lung cancer growth [14, 15]. In addition, natural products targeting the COX and LOX pathways have also been considered as promising treatments for cancer. In this context, curcumin, resveratrol and lycopene, to name a few, act as chemopreventive agents by targeting the COX-2 pathway; while nordihydroguaiaretic acid and baicalein play a beneficial role by inhibiting LOXs [16]. Although promising, bioavailability and possible toxicity of these natural compounds limit their use in clinics. Moreover, their beneficial effects in lung cancer have not been fully explored. In addition to COX-derived products, cytochrome P450 monooxygenase-derived products contribute to lung cancer growth and progression and genetic and/or pharmacological inhibition of these enzymes prevents lung cancer development [1719].

If eicosanoid products play an important role in tumor progression, genetic variation altering eicosanoid producing enzyme function and/or expression may influence NSCLC survival. To address this question, we evaluated the effect of genetic variation within 20 genes that encode eicosanoid metabolism enzymes (of which only 19 had variation) on NSCLC survival in the Southern Community Cohort Study (SCCS) (Fig 1B, S1 Table). In addition, demographic features, such as race and sex, associate with NSCLC survival and may potentially influence genetic associations [5, 2022]. Therefore, we evaluated the effect of interactions between demographic features and genetic variants in eicosanoid metabolism enzymes.


Characterization of cohort

The SCCS is a prospective cohort in the southeastern United States of more than 80,000 individuals that investigates chronic diseases in low income individuals, including non-small cell lung cancer (NSCLC). Participants were enrolled from March 2002 to September 2009, with most participants enrolled at community health centers. A random sample of participants from the general population was selected from voter registration records, driver license records, and commercial mailings. All participants were between the ages of 40 and 79 years at the time of study enrollment. Demographic characteristics of participants recruited at community health centers were obtained during an in-person interview, including self-reported race/ethnicity and tobacco use, while this same information was obtained through mailed questionnaires to participants from the general population. Written informed consent was obtained from all participants. Further details of the SCCS can be found elsewhere [23, 24]. This work was approved by the Vanderbilt University IRB.

Incident NSCLC cases with smoking history were identified through linkage with the 12 state cancer registries overlapping with the SCCS geographic region from 2002 to 2010. From these state cancer registries NSCLC staging and treatment information was abstracted. Staging information was based on American Joint Committee on Cancers (AJCC, 6th and 7th Editions) TNM System staging guidelines [25]. Stage was defined as localized, regional, and distant NSCLC disease with localized disease consisting of AJCC stage I, regional disease consisting of AJCC stages II and III, and distant disease consisting of AJCC stage IV. Treatment information was broadly captured as surgical resection, chemotherapy, or radiation. Vital status was obtained from linkages with the Social Security Administration or the National Death Index as of December 31, 2011. Survival time was defined as the time from date of NSCLC diagnosis to date of death or loss to follow-up. Since certain clinical and demographic variables, such as sex, smoking history, NSCLC stage, and NSCLC treatment may influence NSCLC outcome, the effect of these variables on mortality was examined, coding each as a discrete variable (Table 1). The Cox proportional hazards model was used to determine whether these variables were associated with mortality among non-small cell lung cancer cases in univariate models (Table 2). If a clinical or demographic variable was associated with NSCLC prognosis (p-value < 0.05), the variable was adjusted for in subsequent genetic association models.

Table 1. Description of clinical and demographic variables in NSCLC cases from the SCCS cohort.

Table 2. Univariate association of clinical and demographic variables with mortality in NSCLC cases.

Genotype information

NSCLC cases were genotyped on the Illumina HumanExome BeadChip v1.1. Quality control measures at the individual level excluded any individual with sex inconsistencies or genotyping efficiency less than 98%. Using PLINK, individuals were examined for genetic relatedness; in this sample one pair of relatives was identified and the individual with higher genotyping efficiency was kept in the sample [26]. Based on these criteria our study population consisted of 395 incident NSCLC cases with 265 African Americans and 130 European Americans. SNP level quality control removed SNPs on the sex chromosomes and SNPs < 98% genotyping efficiency.

We extracted genetic information on 20 eicosanoid metabolism genes and examined 108 variants in these genes for association with mortality in NSCLC (Fig 1B, S1 Table). Genomic locations were obtained from the NCBI Gene ID for each gene in GRCh37.p5, and single nucleotide polymorphisms (SNPs) corresponding to the genomic location of each gene were extracted from the larger exome data. For analysis, SNP classification was based on the minor allele frequency (MAF) with rare defined in our study as less than or equal to 10% and common greater than 10%. A relatively high MAF was chosen to define SNPs as rare due to the small sample size. Common and rare SNPs were analyzed independently of each other using different statistical methods.

Common genetic variant analysis

As common SNPs often rely on linkage disequilibrium (LD) with causative variants and LD patterns differ with race/ethnicity, common SNP analyses were stratified by race [27, 28]. The Kaplan-Meier estimator was used to determine which common SNPs were nominally associated with NSCLC mortality (p-value < 0.1). For all survival analyses, censoring occurred at time of death, loss to follow-up, or at the end of the follow-up period (December 31, 2011). As the Kaplan-Meier estimator is a non-parametric statistic, it was used initially to identify genetic variants for further analyses [29]. A Cox proportional hazard model was used to assess the genetic associations for nominally associated SNPs, adjusted for the variables noted above that were significantly associated with prognosis. The effect of common genetic variation was measured using an additive genetic model, except when fewer than 5 homozygotes for the minor allele were present. In this situation, a dominant model was used where the rare homozygotes for the minor allele were combined with the heterozygote individuals. SNP-sex interactions were assessed in African Americans to determine when appropriate to stratify by sex (p-value < 0.05). However, European American SNP-sex interactions were not evaluated due to their small sample size and thus none of these SNP associations were sex stratified. To correct for multiple testing, a false discovery rate (FDR) procedure was used with a FDR of 15% based on the 16 SNPs [30].

Rare genetic variant analysis

Assessment of the effect of rare SNPs on lung cancer survival occurred at the gene-level using the Combined Multivariate and Collapsing (CMC) method [31]. In brief, the CMC method collapses rare genetic variants within a single gene into a binary variable for association analysis [31]. This gene-level variable indicates the presence or absence of rare SNPs within the gene. Since most of the rare SNPs on the genotyping platform used reside in exons, only exonic SNPs composed the gene-level variable. Genes with fewer than 5 individuals possessing a rare SNP within that gene were excluded from analysis (CYP4F2).

We used the Kaplan-Meier estimator to determine the association of the rare gene-level variable with NSCLC survival. Genes associated with NSCLC survival at a p-value < 0.1 were further examined using Cox proportional hazard models; one model was unadjusted, while the other model was adjusted for variables significant in preliminary demographic analyses. To determine when appropriate to stratify by sex for genetic associations, the significance of an interaction term between the collapsed genetic variants and sex was examined in a Cox proportional hazards model. When the p-value was less than 0.05, downstream analyses were conducted stratified by sex. We assessed race and SNP interactions in the same manner to determine when appropriate to stratify by race. Multiple testing was corrected with a FDR of 15% based on the 18 genes tested.

Generation of principal components

Despite examining race-specific genetic interactions for rare variants and stratifying by self-reported race, we cannot exclude the possibility of confounding by residual ancestral population differences in our sample population. To account for any residual ancestral population differences, we generated the first three principal components (PCs) for all individuals using the entire NSCLC population and genome-wide SNPs in EIGENSTRAT [32]. The effect of SNPs significantly associated with lung cancer was calculated in an adjusted model that included other prognostic variables and the first three PCs.

Putative functional effects of SNPs were determined using SIFT, which predicts the effect of amino acid substitutions based on conservation across species [33]. LD plots were generated using r2 values for all SNPs within a given gene using Haploview software [34]. STATA 12.0 (College Station, Texas) was used for all statistical analyses.


From the years of 2002 to 2010, a total of 395 incident cases of NSCLC with smoking history were diagnosed in the SCCS cohort (Table 1). At presentation of disease, 55.2% of the NSCLC cases were diagnosed with distant stage disease. Histology was predominately adenocarcinoma (37.0%), followed by squamous cell carcinoma and NSCLC not otherwise specified (27.9% and 27.6%, respectively). Treatment for NSCLC broadly consisted of surgical resection, chemotherapy, and radiation. Distribution of treatments reflects the restriction of a given treatment to certain stages of NSCLC with no single treatment received by the majority of the population. For example, surgical resection is typically limited to local and regional stage disease. For individuals who underwent the above treatments a combination of chemotherapy and radiation was the most common treatment (21.3%), which is consistent with the frequency of advanced stage disease in this population. Of the NSCLC cases, 22% did not receive any of these three types of treatments. Finally, approximately 15% of the individuals were missing information for one or more treatment variables (Table 3).

Clinical and demographic variables were examined for association with NSCLC mortality (Table 2). Female sex and surgical resection were associated with improved outcome, while increased cigarette consumption was associated with increased mortality. In addition, NSCLC cancer stage associated with survival, with distant disease exhibiting the greatest mortality. The four significantly associated variables, sex, surgical resection, cigarettes per day, and NSCLC stage were included as covariates in all adjusted Cox models.

A total of 108 SNPs in 19 genes were available for analysis (Fig 1B and S1 Table). Common SNP analyses included 16 SNPs in 10 genes (MAF ≥ 0.10) (Fig 1B). After restriction to exonic SNPs, 92 rare SNPs in 18 genes remained for gene-level analyses (Fig 1B). Most of these genes contained between 1 and 5 rare variants within the collapsed variable, with TBXAS1 having the most variants with 13 (Fig 1B). In addition, the COX pathway branch contained the majority of the SNPs with a total of 48 common and rare SNPs (Fig 1B). Of the 19 included genes, only 9 genes were included in both common and rare genetic variant analysis (Fig 1B).

A common SNP in the prostaglandin F synthase (AKR1C3) associates with differences in NSCLC survival in African Americans

Common eicosanoid metabolism gene SNP analyses were stratified by race to account for underlying LD pattern and MAF differences. In European Americans, a total of 12 SNPs in 9 genes and in African Americans a total of 14 SNPs in 9 genes were examined. Of these SNPs, 10 overlapped between the two ethnicities; 2 SNPs were examined exclusively in European Americans and 4 SNPs were examined exclusively in African Americans. SNP-sex interactions were characterized in the African Americans to determine if differential effects were present in the two sexes and if stratification by sex was appropriate for further analyses. Two SNPs, rs2105450 (AKR1C3, prostaglandin F synthase) and rs4792147 (ALOX15B, arachidonate 15-lipoxygenase, type B) had significant SNP-sex interactions with p-values < 0.05 (Table 4), but neither of these SNPs were significantly associated with NSCLC survival in multivariable models among men and women. (S2A–S2D Fig and S2B Table).

Table 4. Common SNPs interaction with sex in African Americans unadjusted for covariates, N = 265.

A single SNP in AKR1C3, rs12529, was associated with NSCLC outcome in African Americans (S3 Table). When adjusting for other prognostic variables, the minor allele of rs12529 (G) remained associated with reduced mortality in African Americans, meaning each additional copy of minor allele for this SNP improved NSCLC survival further (HR = 0.74, 95% CI = 0.59–0.92, p = 0.008) (Table 5 and Fig 5). Adjustment for the first three PCs in addition to the other prognostic variables, revealed this SNP remained significantly associated (p-value = 0.014) (S3 Fig). Of note, rs12529 was not associated with mortality in the European American NSCLC cases despite a MAF greater than 40%, indicating the lack of association is likely not the result of low MAF (S3 and S4 Tables).

Fig 5. Association of a common AKR1C3 SNP, rs12529, with NSCLC survival in African Americans, (N = 265).

Survival curves for each genotypic class are plotted based on the multivariate Cox proportional model, which included adjustment for sex, cigarettes per day, resection, and NSCLC stage.

Table 5. Association of common variants adjusted for prognostic variables.

In addition, one SNP in ALOX15B, rs7225107, was associated with NSCLC prognosis in African Americans, while another ALOX15B SNP, rs9895916, was associated in European Americans (S3 Table). When adjusted for other prognostic variables, neither ALOX15B SNP remained significantly associated with mortality within each race (Table 5). It is noted that ALOX15B SNPs, both rare and common, had extremely low or no LD in both European Americans and African Americans in our sample, so the association of these common SNPs is likely independent from the other genotyped markers (S4 Fig). This pattern of LD is consistent with data from existing databases such as HapMap [27, 28].

Rare exonic variants in ALOX15B Associate with NSCLC prognosis in women

Rare exonic variants in eicosanoid metabolism genes were tested for association with NSCLC survival. Prior to determining the genetic association, the statistical significance of both gene-sex and gene-race interactions was examined to determine when stratifying by sex and race, respectively, was appropriate for evaluating the genetic association. Two of these interactions were significant: the interaction between ALOX15B and sex and TBXAS1 and race (S5 Table). However, rare variants in TBXAS1 did not associate with NSCLC survival in either race stratified.

Initial associations examined using the Kaplan-Meier estimator indicated that none of the rare variants in eicosanoid metabolism genes were associated with differences in NSCLC survival, but the presence of the interaction between ALOX15B and sex suggested the effect of rare variants in ALOX15B on prognosis differs by sex (S5 Table). Sex-stratified analyses showed rare ALOX15B exonic variants were associated with decreased survival in men, but associated with increased survival in women, although the association was only statistically significant among women (Fig 6). To further characterize this association in women, the association of ALOX15B variants was adjusted for surgical resection, cigarettes per day, and NSCLC staging. The association of ALOX15B variants remained statistically significant in the adjusted model (HR = 2.10, 95% CI = 1.25–3.54, p-value = 0.005) (Fig 7). Inclusion of the first three PCs, in addition to the other prognostic variables, did not appreciably alter this association (HR = 2.07) (S5 Fig). Finally, although rare exonic variants in ALOX12 and PTGIS were associated with NSCLC mortality (p-value < 0.05) using the log-rank test these associations were not statistically significant in adjusted models (S6 and S7 Tables).

Fig 6. Rare Exonic ALOX15B SNPs associate with a worse prognosis in female NSCLC patients.

Sex-stratified survival curves calculated from the corresponding sex-specific univariate Cox proportional hazard model are shown above with the hazard ratio and p-values reported from the Cox proportional hazard model. The blue survival line is individuals without any rare ALOX15B variants and the red survival line is individuals with rare ALOX15B variants. (A) The survival curve and association of ALOX15B SNPs in men (N = 217), and (B) the survival curve and association in women (N = 175).

Fig 7. Rare ALOX15B SNPs associate with worse prognosis in adjusted analysis, (N = 175).

Multiple variable adjusted survival curves for women were generated from an adjusted Cox proportional model. Variables adjusted for include resection, cigarettes per day, and NSCLC staging. Both the hazard ratio and p-value from the Cox proportional hazard model are reported.


The goal of this study was to determine whether exonic variation in the eicosanoid pathway is associated with NSCLC survival. To do this, we investigated the association of genetic variation in eicosanoid metabolism genes with NSCLC prognosis in the Southern Community Cohort Study. No genetic variants were significantly associated with NSCLC mortality. The lack of an association for the entire population is not unexpected since prognosis differs by sex. However, variation within ALOX15B and the prostaglandin F synthase AKR1C3 associated significantly with NSCLC survival in certain subgroups of the NSCLC cases. We found that rare exonic ALOX15B variants were associated with poorer survival among women with NSCLC and rs12529 in AKR1C3 was associated with better prognosis in African Americans with NSCLC. Despite the decreased survival in women with rare ALOX15B variants, overall women still survive longer than men. This may be because the SNPs we studied were too rare to affect outcome averaged across all women. Although this may also be due to women being more proactive in their treatment, data indicates that under equal treatment women still survive better than men indicating some still unknown factors impacting survival [3537].

Our results show an independent association between sex and survival with female NSCLC cases surviving longer than men with the same disease. This is consistent with prior studies also demonstrating improved NSCLC survival in women [3537]. Some studies indicate that better NSCLC survival in women is limited to or more pronounced in cases with adenocarcinoma and this histology is more common in women, but other studies have failed to replicate these findings [3539] The reasons for the differential survival between men and women are not well understood; however, one study suggested that tumors from women with NSCLC are enriched for changes in expression of immunity-related genes despite having similar composition of immune cells to their male counterparts [40]. Along with our own findings, these studies indicate that NSCLC may have important biological differences between the two sexes that drive survival differences.

ALOX15B, or 15-LOX-2, is a lipoxygenase enzyme that metabolizes arachidonic acid to produce 15(S)-HETE. ALOX15B has been proposed as a tumor suppressor gene, as its major product 15(S)-HETE inhibits prostate cancer cell cycle progression and VEGF synthesis [4143]. In contrast to these studies, 15(S)-HETE promotes angiogenesis in adipose tissues and stimulates VEGF production in endothelial cells derived from adipose tissues, suggesting that this lipid mediator can either prevent or promote angiogenesis and that its effects are cell type and tissue specific [44]. The role of ALOX15B in lung cancer biology is poorly characterized. Although this enzyme is expressed in both normal and cancerous lung tissues, ALOX15B expression correlates inversely with both NSCLC cancer cell proliferation and NSCLC cancer grade [45, 46]. The inverse correlation between ALOX15B levels and lung cancer grade can be explained by the fact that 15(S)-HETE can bind and activate PPARγ, a nuclear transcription factor shown to inhibit the growth of human NSCLC cells as well as prevent smoking-induced lung cancer development in mice [45, 47, 48]. Interestingly, half of the associated rare exonic ALOX15B SNPs we have identified are predicted to be loss-of-function by SIFT (S8 Table). Thus, the worse prognosis in NSCLC individuals is consistent with the idea that ALOX15B-derived 15(S)-HETE plays a protective role in the course of lung cancer either by activating PPARγ or by downregulating the expression of pro-angiogenic genes.

AKR1C3 acts on a large number of substrates including steroid hormones, prostaglandins, and proximate carcinogens of polycyclic aromatic hydrocarbons (PAHs) [4955]. Among its enzymatic activity AKR1C3 generates PGF2 epimers that bind and activate prostaglandin F receptors [56, 57]. Increased PGF production in endometrial cancer promotes progression by enhancing angiogenesis and proliferation [5861]. The potential role of PGF2 epimers in NSCLC biology is not well characterized, but given their role in promoting endometrial tumorigenesis it is certainly plausible these molecules have similar effects in NSCLC, and prior genetic studies have identified AKR1C3 as associating with bladder and lung cancer risk [62, 63]. Here we found that a common variant in AKR1C3, rs12529, associated with improved NSCLC prognosis in African Americans. The failure to find a comparable trend in the European descent populations cannot be explained at the present, especially as we do not know the exact nature of the causal relationship even in African Americans.

Mechanistically, the observed association of rs12529 may be explained several ways. The associated SNP may in itself be causative, but it should be noted that while rs12529 is located within the coding region of AKR1C3 SIFT predicts the SNP as benign and no studies have characterized this SNP as functional to date. In addition, although AKR1C3 possesses prostaglandin-F synthase activity, its ability to act on other classes of substrate cannot be excluded as a mechanistic explanation for the observed association with NSCLC survival. Another explanation for this association is that rs12529 associates indirectly with NSCLC prognosis, meaning rs12529 is in LD with a causative genetic variation rather than being causative itself. Further characterization of genetic variation in AKR1C3 will be necessary to distinguish these possibilities.

Despite the two associations found above, there are limitations to our study. First, the sample size of this study was relatively small, especially when examining associations among strata. It is likely that even if these associations are real that the effect size is inaccurate creating a need to replicate these findings to better estimate the effect size. However, replication was not attempted here due to the scarcity of an appropriate population for replication. Another limitation was the use of the CMC method for analysis of rare variants in the event the rare variants have opposite directions of effect, biasing the association towards the null. In addition, the CMC method cannot capture differences in effect size among rare variants. As a result the association of collapsed rare variants represents an average effect size that may misestimate the real effects [31].

The association of arachidonic acid genetic variation with NSCLC survival suggests NSCLC cases with certain genotypes may benefit from treatments targeting eicosanoid metabolism, and that patient attributes, such as genotype, sex, and race, may also influence NSCLC treatment decisions. We found multiple sex-SNP interactions that influence NSCLC survival, where sex determined the direction of effect for a SNP. On this basis, treatments targeting eicosanoid metabolism will likely have sex-specific differences. However, we cannot rule out the role of sex-specific NSCLC histology differences in driving sex-SNP interactions due to inadequate sample size. Despite these limitations, the association of rare variants in ALOX15B in women may begin to explain the observed differences in NSCLC survival between men and women. Overall, these findings still suggest a more personalized approach in NSCLC treatment would benefit patients and targeting eicosanoid metabolism may be a viable NSCLC treatment in certain patients.

In conclusion, we have identified two arachidonic acid metabolizing enzymes whose genetic variation associated significantly with NSCLC survival. Their effect appears to differ by sex and race, indicating that different approaches to treatment may be necessary as a result of either. Therefore, it is possible that eicosanoid metabolism interventions may be used to slow NSCLC progression. Finally, since the pathways we have studied have existing drugs to treat cardiovascular disease, a number of drugs already exist that may be potentially re-purposed for the treatment of NSCLC.

Supporting information

S2 Fig. Common SNPs in arachidonic acid metabolism genes interact with sex to influence African American NSCLC progression


S3 Fig. The association of rs12529 with NSCLC survival in African Americans, n = 265.


S5 Fig. Rare ALOX15B variants associate with worse prognosis when adjusted for variables including top 3 PCs, n = 175.


S1 Table. Arachidonic acid metabolism pathway genes.


S2 Table. Association of SNPs interacting with sex in African Americans.


S3 Table. Association of common SNPs within eicosanoid metabolism genes.


S4 Table. Comparison of minor allele frequencies in AKR1C3 SNPs in European and African Americans.


S5 Table. Interaction of demographic variables with rare genetic variants in NSCLC survival association, n = 395.


S6 Table. Association of rare SNPs collapsed by gene with NSCLC survival, n = 395.


S7 Table. Univariate and multivariate associations of rare SNPs.


S8 Table. SIFT predictions for rare ALOX15B variants.



  1. 1. Sher T, Dy GK, Adjei AA. Small cell lung cancer. Mayo Clinical Processions. 2008;83:355–267.
  2. 2. Torre LA, Bray F, Siegel RL, Ferlay J, Lortet-Tieulent J, Jemal A. Global Cancer Statistics, 2012. CA: Cancer Journal for Clinicans. 2015;65:87–108.
  3. 3. Siegel R, Naishadham D, Jemal A. Cancer statistics, 2013. CA: Cancer Journal for Clinicans. 2013;63:11–30.
  4. 4. Society AC. Cancer Facts & Figures. 2013.
  5. 5. Howlader N, Noone AM, Krapcho M, Garshell J, Neyman N, Altekruse SF, et al. SEER Cancer Statistics Review, 1975–2010. 2013.
  6. 6. Hanahan D, Weinberg RA. The hallmarks of cancer. Cell. 2000;100:57–70. pmid:10647931
  7. 7. Schneider C, Pozzi A. Cyclooxygenases and lipoxygenases in cancer. Cancer metastasis reviews. 2011;30(3–4):277–94. Epub 2011/10/18. pmid:22002716; PubMed Central PMCID: PMC3798028.
  8. 8. Eberhart CE, Coffey RJ, Radhika A, Giardiello FM, Ferrenbach S, DuBois RN. Up-regulation of cyclooxygenase 2 gene expression in human colorectal adenomas and adenocarcinomas. Gastroenterology. 1994;107(4):1183–8. Epub 1994/10/01. pmid:7926468.
  9. 9. Soslow RA, Dannenberg AJ, Rush D, Woerner BM, Khan KN, Masferrer J, et al. COX-2 is expressed in human pulmonary, colonic, and mammary tumors. Cancer-Am Cancer Soc. 2000;89(12):2637–45. Epub 2001/01/03. pmid:11135226.
  10. 10. Wang D, Dubois RN. Eicosanoids and cancer. Nat Rev Cancer. 2010;10(3):181–93. Epub 2010/02/20. pmid:20168319; PubMed Central PMCID: PMC2898136.
  11. 11. Achiwa H, Yatabe Y, Hida T, Kuroishi T, Kozaki K, Nakamura S, et al. Prognostic significance of elevated cyclooxygenase 2 expression in primary, resected lung adenocarcinomas. Clinical cancer research: an official journal of the American Association for Cancer Research. 1999;5(5):1001–5. Epub 1999/06/03. pmid:10353732.
  12. 12. Khuri FR, Wu H, Lee JJ, Kemp BL, Lotan R, Lippman SM, et al. Cyclooxygenase-2 overexpression is a marker of poor prognosis in stage I non-small cell lung cancer. Clinical cancer research: an official journal of the American Association for Cancer Research. 2001;7(4):861–7. Epub 2001/04/20. pmid:11309334.
  13. 13. Brabender J, Park J, Metzger R, Schneider PM, Lord RV, Holscher AH, et al. Prognostic significance of cyclooxygenase 2 mRNA expression in non-small cell lung cancer. Annals of surgery. 2002;235(3):440–3. Epub 2002/03/08. pmid:11882767; PubMed Central PMCID: PMC1422451.
  14. 14. Zhang MZ, Xu J, Yao B, Yin H, Cai Q, Shrubsole MJ, et al. Inhibition of 11beta-hydroxysteroid dehydrogenase type II selectively blocks the tumor COX-2 pathway and suppresses colon carcinogenesis in mice and humans. The Journal of clinical investigation. 2009;119(4):876–85. Epub 2009/03/25. pmid:19307727; PubMed Central PMCID: PMC2662561.
  15. 15. Chang J, Xue M, Yang S, Yao B, Zhang B, Chen X, et al. Inhibition of 11beta-Hydroxysteroid Dehydrogenase Type II Suppresses Lung Carcinogenesis by Blocking Tumor COX-2 Expression as Well as the ERK and mTOR Signaling Pathways. Plos One. 2015;10(5):e0127030. Epub 2015/05/27. pmid:26011146; PubMed Central PMCID: PMC4444260.
  16. 16. Yarla NS, Bishayee A, Sethi G, Reddanna P, Kalle AM, Dhananjaya BL, et al. Targeting arachidonic acid pathway by natural products for cancer prevention and therapy. Seminars in cancer biology. 2016. Epub 2016/02/09. pmid:26853158.
  17. 17. Pozzi A, Ibanez MR, Gatica AE, Yang S, Wei S, Mei S, et al. Peroxisomal proliferator-activated receptor-alpha-dependent inhibition of endothelial cell proliferation and tumorigenesis. J Biol Chem. 2007;282(24):17685–95. Epub 2007/04/05. pmid:17405874.
  18. 18. Pozzi A, Popescu V, Yang SL, Mei SJ, Shi MJ, Puolitaival SM, et al. The Anti-tumorigenic Properties of Peroxisomal Proliferator-activated Receptor alpha Are Arachidonic Acid Epoxygenase-mediated. J Biol Chem. 2010;285(17):12840–50. pmid:20178979
  19. 19. Skrypnyk N, Chen XW, Hu W, Su Y, Mont S, Yang SL, et al. PPAR alpha Activation Can Help Prevent and Treat Non-Small Cell Lung Cancer. Cancer Res. 2014;74(2):621–31. pmid:24302581
  20. 20. Kawaguchi T, Takada M, Kubo A, Matsumura A, Fukai S, Tamura A, et al. Performance status and smoking status are independent favorable prognostic factors for survival in non-small cell lung cancer: a comprehensive analysis of 26,957 patients with NSCLC. Journal of thoracic oncology: official publication of the International Association for the Study of Lung Cancer. 2010;5(5):620–30. Epub 2010/04/01. pmid:20354456.
  21. 21. Brundage MD, Davies D, Mackillop WJ. Prognostic factors in non-small cell lung cancer: a decade of progress. Chest. 2002;122(3):1037–57. Epub 2002/09/13. pmid:12226051.
  22. 22. Nakamura H, Ando K, Shinmyo T, Morita K, Mochizuki A, Kurimoto N, et al. Female gender is an independent prognostic factor in non-small-cell lung cancer: a meta-analysis. Annals of Thoracic and Cardiovascular Surgery. 2011;17:469–80. pmid:21881356
  23. 23. Signorello LB, Hargreaves MK, Blot WJ. The Southern Community Cohort Study: investigating health disparities. Journal Health Care Poor Underserved. 2010;21:26–7.
  24. 24. Signorello LB, Hargreaves MK, Steinwandel MD, Zheng W, Cai Q, Schlundt DG, et al. Southern community cohort study: establishing a cohort to investigate health disparities. Journal of National Medical Association 2005;97:972–9.
  25. 25. Aldrich MC, Grogan EL, Munro HM, Signorello LB, Blot WJ. Stage-adjusted Lung Cancer Survival Does Not Differ Between Low-Income Blacks and Whites. Journal of Thoracic Oncology 2013;8:1248–54. pmid:24457235
  26. 26. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet. 2007;81(3):559–75. Epub 2007/08/19. pmid:17701901; PubMed Central PMCID: PMC1950838.
  27. 27. International HapMap C. A haplotype map of the human genome. Nature. 2005;437(7063):1299–320. Epub 2005/10/29. pmid:16255080; PubMed Central PMCID: PMC1880871.
  28. 28. International HapMap C, Frazer KA, Ballinger DG, Cox DR, Hinds DA, Stuve LL, et al. A second generation human haplotype map of over 3.1 million SNPs. Nature. 2007;449(7164):851–61. Epub 2007/10/19. pmid:17943122; PubMed Central PMCID: PMC2689609.
  29. 29. Kaplan EL, Meier P. Nonparametric-Estimation from Incomplete Observations. J Am Stat Assoc. 1958;53(282):457–81.
  30. 30. Benjamini Y, Drai D, Elmer G, Kafkafi N, Golani I. Controlling the false discovery rate in behavior genetics research. Behavioural brain research. 2001;125(1–2):279–84. Epub 2001/10/30. pmid:11682119.
  31. 31. Li B, Leal SM. Methods for Detecting Association with Rare Variants for Common Disease: Application to Analysis of Sequence Data. American journal of human genetics. 2008;83:311–21. pmid:18691683
  32. 32. Price AL, Patterson NJ, Plenge RM, Weinblatt ME, Shadick NA, Reich D. Principal components analysis corrects for stratification in genome-wide association studies. Nature genetics. 2006;38(8):904–9. pmid:16862161
  33. 33. Ng PC, Henikoff S. SIFT: Predicting amino acid changes that affect protein function. Nucleic acids research. 2003;31(13):3812–4. Epub 2003/06/26. pmid:12824425; PubMed Central PMCID: PMC168916.
  34. 34. Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics. 2005;21(2):263–5. Epub 2004/08/07. pmid:15297300.
  35. 35. Batevik R, Grong K, Segadal L, Stangeland L. The female gender has a positive effect on survival independent of background life expectancy following surgical resection of primary non-small cell lung cancer: a study of absolute and relative survival over 15 years. Lung cancer. 2005;47(2):173–81. Epub 2005/01/11. pmid:15639716.
  36. 36. Wheatley-Price P, Blackhall F, Lee SM, Ma C, Ashcroft L, Jitlal M, et al. The influence of sex and histology on outcomes in non-small-cell lung cancer: a pooled analysis of five randomized trials. Annals of oncology: official journal of the European Society for Medical Oncology / ESMO. 2010;21(10):2023–8. Epub 2010/03/25. pmid:20332134.
  37. 37. Sagerup CM, Smastuen M, Johannesen TB, Helland A, Brustugun OT. Sex-specific trends in lung cancer incidence and survival: a population study of 40,118 cases. Thorax. 2011;66(4):301–7. Epub 2011/01/05. pmid:21199818.
  38. 38. Visbal AL, Williams BA, Nichols FCr, Marks RS, Jett JR, Aubry MC, et al. Gender differences in non-small-cell lung cancer survival: an analysis of 4,618 patients diagnosed between 1997 and 2002. The Annals of Thoracic Surgery. 2004;78:209–15. pmid:15223430
  39. 39. de Perrot M, Licker M, Bouchardy C, Usel M, Robert J, Spiliopoulos A. Sex differences in presentation, management, and prognosis of patients with non-small cell lung carcinoma. The Journal of thoracic and cardiovascular surgery. 2000;119(1):21–6. Epub 1999/12/29. pmid:10612756.
  40. 40. Araujo JM, Prado A, Cardenas NK, Zaharia M, Dyer R, Doimi F, et al. Repeated observation of immune gene sets enrichment in women with non-small cell lung cancer. Oncotarget. 2016. Epub 2016/03/10. pmid:26958810.
  41. 41. Tang S, Bhatia B, Maldonado CJ, Yang P, Newman RA, Liu J, et al. Evidence that arachidonate 15-lipoxygenase 2 is a negative cell cycle regulator in normal prostate epithelial cells. J Biol Chem. 2002;277:16189–201. pmid:11839751
  42. 42. Bhatia B, Maldonado CJ, Tang S, Chandra D, Klein RD, Chopra D, et al. Subcellular localization and tumor-suppressive functions of 15-lipoxygenase 2 (15-LOX2) and its splice variants. J Biol Chem. 2003;278:25091–100. pmid:12704195
  43. 43. Tang Y, Wang MT, Chen Y, Yang D, Che M, Honn KV, et al. Downregulation of vascular endothelial growth factor and induction of tumor dormancy by 15-lipoxygenase-2 in prostate cancer. International Journal of Cancer. 2009;124:1545–51. pmid:19089921
  44. 44. Soumya SJ, Binu S, Helen A, Reddanna P, Sudhakaran PR. 15(S)-HETE-induced angiogenesis in adipose tissue is mediated through activation of PI3K/Akt/mTOR signaling pathway. Biochemistry and cell biology = Biochimie et biologie cellulaire. 2013;91(6):498–505. Epub 2013/11/14. pmid:24219292.
  45. 45. Yuan H, Li M, Ma L, Hsin M, Mok TSK, Underwood MJ, et al. 15-Lipoxygenases and its metabolites 15(S)-HETE and 13(S)-HODE in the development of non-small cell lung cancer. Thorax. 2010;65:321–6. pmid:20388757
  46. 46. Gonzalez AL, Roberts RL, Massion PP, Olson SJ, Shyr Y, Shappell SB. 15-Lipoxygenase-2 Expression in benign and Neoplastic Lung: An Immunohistochemical Study and Correlation With Tumor Grade and Proliferation. Human Pathology. 2004;35:840–9. pmid:15257547
  47. 47. Keshamouni VG, Reddy RC, Arenberg DA, Joel B, Thannickal VJ, Kalemkerian GP, et al. Peroxisome proliferator-activated receptor-gamma activation inhibits tumor progression in non-small-cell lung cancer. Oncogene. 2004;23:100–8. pmid:14712215
  48. 48. Li MY, Kong AW, Yuan H, Ma LT, Hsin MK, Wan IY, et al. Pioglitazone prevents smoking carcinogen-induced lung tumor development in mice. Current cancer drug targets. 2012;12(6):597–606. Epub 2012/04/03. pmid:22463586.
  49. 49. Lin HK, Jez JM, Schlegel BP, Peehl DM, Pachter JA, Penning TM. Expression and characterization of recombinant type 2 3a-hydroxysteroid dehydrogenase (HSD) from human prostate: demonstration of bifunctional 3a/17b-HSD activity and cellular distribution. Molecular Endocrinology. 1997;11:1971–84. pmid:9415401
  50. 50. Byrns MC, Duan L, Lee SH, Blair IA, Penning TM. Aldo-keto reductase 1C3 in MCF-7 cells reveals roles in steroid hormone and prostaglandin metabolism that may explain its overexpression in breast cancer. The Journal of steroid biochemistry and molecular biology. 2010;118:177–87. pmid:20036328
  51. 51. Suzuki-Yamamoto T, Nishizawa M, Fukui M, Okuda-Ashitaka E, Nakajima T, Ito S, et al. cDNA cloning, expression and characterization of human prostaglandin F synthase. FEBS Lett. 1999;462(3):335–40. Epub 2000/01/06. pmid:10622721.
  52. 52. Koda N, Tsutsui Y, Niwa H, Ito S, Woodward D, Watanabe K. Synthesis of prostaglandin F ethanolamide by prostaglandin F synthase and identification of bimatoprost as a potent inhibitor of the enzyme: new enzyme assay method using LC/ESI/MS. Archives of Biochemistry and Biophysics 2004;424:128–36. pmid:15047184
  53. 53. Penning TM, Ohnishi ST, Ohnishi T, Harvey RG. Generation of reactive oxygen species during the enzymatic oxidation of polycyclic aromatic hydrocarbon trans-dihydrodiols catalyzed by dihydrodiol dehydrogenase. Chemical Research in Toxicology 1996;9:84–92. pmid:8924621
  54. 54. Shou M, Harvey RG, Penning TM. Reactivity of benzo[a]pyrene-7,8-dione with DNA. Evidence for the formation of deoxyguanosine adducts. Carcinogenesis. 1993;14:475–82. pmid:8384091
  55. 55. Palackal NT, Lee SH, Harvey RG, Blair IA, Penning TM. Activation of polycyclic aromatic hydrocarbon trans-dihydrodiol proximate carcinogens by human aldo-keto reductase (AKR1C) enzymes and their functional overexpression in human lung carcinoma (A549) cells. J Biol Chem. 2002;277:24799–808. pmid:11978787
  56. 56. Shigematsu H, Lin L, Takahashi T, Nomura M, Suzuki M, Wistuba II, et al. Clinical and biological features associated with epidermal growth factor receptor gene mutations in lung cancers. J Natl Cancer Inst. 2005;97(5):339–46. Epub 2005/03/03. pmid:15741570.
  57. 57. Abramovitz M, Boie Y, Nguyen T, Rushmore TH, Bayne MA, Metters KM, et al. Cloning and expression of a cDNA for the human prostanoid FP receptor. J Biol Chem. 1994;269(4):2632–6. Epub 1994/01/28. pmid:8300593.
  58. 58. Keightley MC, Sales KJ, Jabbour HN. PGF2α-F-prostanoid receptor signaling via ADAMTS1 modulates epithelial cell invasion and endothelial cell function in endometrial cancer. BMC Cancer. 2010;10:488. pmid:20840749
  59. 59. Wallace AE, Catalano RD, Anderson RA, Jabbour HN. Chemokine (C-C) motif ligand 20 is regulated by PGF(2α)-F-prostanoid receptor signalling in endometrial adenocarcinoma and promotes cell proliferation. Molecular and Cellular Endocrinology 2011;331:129–35. pmid:20816914
  60. 60. Keightley MC, Brown P, Jabbour HN, Sales KJ. F-Prostaglandin receptor regulates endothelial cell function via fibroblast growth factor-2. BMC Cell Biology. 2010;11:8. pmid:20092633
  61. 61. Sales KJ, List T, Boddy SC, Williams AR, Anderson RA, Naor Z, et al. A novel angiogenic role for prostaglandin F2alpha-FP receptor interaction in human endometrial adenocarcinomas. Cancer Res. 2005;65:7707–16. pmid:16140938
  62. 62. Lan Q, Mumford JL, Shen M, Demarini DM, Bonner MR, He X, et al. Oxidative damage-related genes AKR1C3 and OGG1 modulate risks for lung cancer due to exposure to PAH-rich coal combustion emissions. Carcinogenesis. 2004;25:2177–81. pmid:15284179
  63. 63. Figueroa JD, Malats N, García-Closas M, Real FX, Silverman D, Kogevinas M, et al. Bladder cancer risk and genetic variation in AKR1C3 and other metabolizing genes. Carcinogenesis. 2008;29:1955–62. pmid:18632753