Common filaggrin gene mutations and risk of cervical cancer.

BACKGROUND
As carriers of filaggrin gene (FLG) mutations may have a compromised cervical mucosal barrier against human papillomavirus infection, our primary objective was to study their risk of cervical cancer.


METHODS
We genotyped 586 cervical cancer patients for the two most common FLG mutations, R501X and 2282del4, using blood from the Copenhagen Hospital Biobank, Denmark. Controls (n = 8050) were genotyped in previous population-based studies. Information on cervical cancer, mortality and emigration were obtained from national registers. Odds ratios (OR) were estimated by logistic regression with adjustment for age at blood sampling, and weighted by the genotype-specific inverse probability of death between diagnosis and sampling. Hazard ratios (HR) were estimated by Cox regression with time since diagnosis as underlying time, and with adjustment for age at diagnosis and stratification by cancer stage.


RESULTS
The primary results showed that FLG mutations were not associated with the risk of cervical cancer (6.3% of cases and 7.7% of controls were carriers; OR adjusted 0.81, 95% CI 0.57-1.14; OR adjusted+ weighted 0.96, 95% CI 0.58-1.57). Among cases, FLG mutations increased mortality due to cervical cancer (HR 4.55, 95% CI 1.70-12.2), however, the association was reduced after stratification by cancer stage (HR 2.53, 95% CI 0.84-7.59).


CONCLUSION
Carriage of FLG mutations was not associated with the risk of cervical cancer.

Filaggrin proteins are essential for the formation and function of the skin but are also expressed in oral and cervical mucosa. The proteins are known to align intracellular fi laments leading to fl attening of surface cell keratinocytes (squamous cells) and the degradation products maintain hydration, low pH, and provide immune modulation [1]. The impact of fi laggrin defi ciency on skin diseases was recognized in 2006, when two loss-of-function mutations in the fi laggrin gene ( FLG ), causing reduced expression of fi laggrin, were discovered and shown to occur in 8 -10% of North European and 5% of Asian study populations. The mutations were detected in almost all patients with the disease ichthyosis vulgaris -a semidominantly inherited disorder characterized by xerosis and palmar hyper linearity -and were further strongly associated with atopic dermatitis [2,3]. FLG mutations are the fi rst strong genetic risk factor yet Acta Oncologica, 2015;54: 217-223 identifi ed in the concept of complex diseases, which also encompass cervical cancer, and the fi rst strong risk factor for atopic dermatitis [4,5].
In the 1980s, the expression of fi laggrin in epithelia from cervical and oral biopsies was studied for its predictive value for premalignant or malignant squamous lesion [6 -12]. In particular, fi laggrin expression was found to be reduced in cervical lesions infected with the carcinogenic human papillomavirus (HPV) in a dose-dependent manner [6 -8]. Currently, the role of fi laggrin in the cervix remains unclear. We hypothesize that in a fi laggrin-reduced cervix, HPV may have increased access and persistence into its target cells, the basal cells, for example due to reduced cervical acidity [13,14], compromised adherence between cells in the upper layers, or ongoing infection [15]. Results of epidemiological studies are compatible with this suspicion. For example, atopic dermatitis, in which up to 40% of patients carry FLG mutations, has been associated with higher occurrence of skin infections with wart viruses (HPV, Molluscum contagiosum ), herpes simplex, herpes zoster, and Staphylococcus aureus [16 -18]. A few of these studies had information on the fi laggrin genotype and showed a higher prevalence of skin infections in mutation carriers with atopic dermatitis [19,20]. Furthermore, a study of two longitudinal cohorts showed an increased risk of cervical cancer in women who had atopic dermatitis and common warts in childhood [21]. As the role of FLG mutations in the etiology of cervical cancer has not been studied previously, we took advantage of the national biobank and patient registers in Denmark to genotype blood for the two most common FLG mutations, R501X and 2282del4, and examine the risk of cervical cancer in FLG mutation carriers.

Study population
Cases and controls were nested in the general population in the capital region in Denmark. Supplementary Table I (available online at http:// informahealthcare.com/doi/abs/10.3109/0284186X .2014.973613) shows the study fl ow. We identifi ed 808 in-or outpatients with a discharge diagnosis of cervical cancer and a blood sample in the Copenhagen Hospital Biobank by person-linkage between the National Patient Registry and the National Biobank Register (www.biobankdenmark.dk) using as link the Civil Registration System (CRS) number assigned to all Danish residents [22]. A total of 778 cases were genotyped; the remaining had no genotype call. The study included 586 cases whose fi rst discharge diagnosis of cervical cancer verifi ed in the Danish Cancer Register updated until 2011 (denoted ' verifi ed cases ' ). The remaining cases were not considered eligible for inclusion, either because the discharge diagnosis was before the update but had never been verifi ed (81 patients), or after the update and verifi cation and also mortality was therefore yet unknown (110 patients). Controls included 8050 genotyped females from four previous populationbased studies, excluding 25 females who developed cervical cancer after the blood sample was drawn.

Cervical cancer
Potential cases of cervical cancer were identifi ed in the National Patient Register updated until May 2013 (ICD-8, code 180; ICD-10, code DC53) [23]. From the Danish Cancer Register updated until 2011 we obtained information on all cancer, morphology, and stage for all women in Denmark with cervical cancer including the studied cases (ICD-7, code 171; ICD-10, code DC53) [24,25]. Vital status and cause-specifi c mortality was obtained from the Civil Registration System and the Danish Register of Causes of Death updated until 2011 (Death from cervical cancer: ICD-8, code 180; ICD-10, code DC53) [22,26].

Control cohorts
Controls participated in four studies described elsewhere: ' MONICA10 ' , 1,291 females [27]; ' Inter99 ' , 3081 females [28]; ' Health 2006 ' , 1848 females [29]; a study of the genetics of preterm delivery, 1855 females [30,31]. Briefl y, the fi rst three studies were general health studies of randomly selected adult men and women in the capital region in 1983 -2008 [32] and the last study was a study including non-complicated pregnancies with preterm (cases) or term (controls) delivery and was nested among pregnant women who enrolled in 1996 -2002 in the large Danish National Birth Cohort study (see also Supplementary Table I available online at http:// informahealthcare.com/doi/abs/10.3109/0284186X. 2014.973613) [33].

Copenhagen Hospital Biobank
The Copenhagen Hospital Biobank was established at the Department of Clinical Immunology at Copenhagen University Hospital, Rigshospitalet, to facilitate easier and standardized access to samples for medical research. Since February 2009, the Copenhagen Hospital Biobank has stored EDTA whole blood in excess from patient samples submitted for blood typing at the Copenhagen University Hospital, Rigshospitalet. Since February 2012, Copenhagen Hospital Biobank stored similar sample excess from all other hospitals in the capital region (except from Copenhagen University Hospital Bornholm and psychiatric hospitals). Patients were included only once, i.e. repeated samples from the same patient were not included. At inclusion, May 2013, the Copenhagen Hospital Biobank contained more than 100 000 samples.

Filaggrin gene (FLG) mutations
For both cases and controls, we obtained information on the two most common null mutations in the FLG among Northern European descendants (GenBank NM_002016.1; c.1537C Ͼ T and c.2318_2321del; located on chromosome 1q21), commonly designated R501X and 2282del4, respectively. For cases and DNBC controls, DNA was genotyped with KASP ™ technology (LGC Genomics Ltd, Herts, UK), whereas DNA from remaining controls were genotyped previously by Multiplex analysis (Department of Clinical Biochemistry at Copenhagen University Hospital, Gentofte, Denmark) [34]. Both methods uses a 3 ' anchored reverse primer to target and distinguish FLG mutations due to their sequence homology. As heterozygous carriers still produce fi laggrin on the competent allele, we categorized zygosity: We defi ned carriers as homozygous when having the same mutation on both alleles, heterozygous when having one mutation only, and compound heterozygous when having different mutations on the two alleles (grouped as homozygous).

Ethics
The study was approved by the Danish Data Protection Agency (j. nr. 2008-54-0472), the steering committee of the Copenhagen Hospital Biobank, and the Ethical Committees of the Capital Region (j.nr. H-1-2013-055). Based on § 10 of the Ethical committee law, dispensation for obtaining informed consent was granted because the studied blood samples were obtained as part of treatment and stored in a research biobank, all data were treated anonymous, results do not occur at the individual level, results of analyses have no consequences for the individual participant, the Danish law on handling of personal information is kept, and participants registered in the Danish Tissue Application Register are excluded.

Statistical analyses
The potential association of FLG mutations with the risk of cervical cancer was analyzed by comparing the occurrence of FLG mutations in cases and controls using an odds ratio (OR) estimated in a logistic regression analysis with adjustment for age at blood sampling (one-year categories). In an additional analysis, the logistic regression was performed as a weighted analysis weighted by the genotype-specifi c inverse probability of survival between diagnoses and sampling. Survival estimates used in the weights were calculated using the Kaplan Meier estimator based on follow-up the fi rst year. ORs for subgroups of cases were estimated by comparing the selected cases with all controls. The potential association of prognosis with FLG mutations among cervical cancer patients was analyzed by comparing mortality rates in patients with and without FLG mutations using a hazard ratio (HR) estimated in a Cox regression with time since diagnosis as underlying time, and with adjustment for age at diagnosis (continuous) and stratifi cation by cancer stage (1, 2, 3, 4, localized, regional). Cervical cancer cases were followed from time of diagnosis of cervical cancer or sampling date, whichever came later (i.e. delayed entry), until end of 2011 or date of death from cervical cancer or other cause, whichever came fi rst. Maximum likelihood estimation of ORs and HRs was performed using the PROC GENMOD and PROC PHREG procedure in SAS, version 9.4 software (SAS Institut, Inc., Cary, NC, USA). 95% CI for odds ratios were based on Wald ' s approximation except for the weighted logistic regression were bootstrapping with 5000 replications was used. Stage-specifi c ORs were compared by multinomial regression using PROC NLMIXED in SAS where a trend test was performed by assigning the values 1, 2, 3, and 4 to cases with stages 1 -4 and treating stage as a continuous variable. Cancer characteristics for the cases in the study and all cervical cancer cases in Denmark were compared by a χ 2 -test, except survival after diagnosis which was compared in a Kaplan-Meier plot. Table I shows the characteristics of the 586 cases and 8050 controls. Overall, age at blood sampling was comparable despite the different years of blood sampling (cases, 2009 -2013; controls, 1993 -2008). However, older cases and younger controls were slightly more frequent (Table I), and two of four control cohorts were sampled with 5-and 10-year age intervals (data not shown). Most cases had blood sampled after being diagnosed with cervical cancer (7 months to 43 years, 53%; 2 -6 months, 10%; Ͻ 2 months, 28%) whereas few cases were sampled before (1%). Cases were comparable with all cervical cancer patients in Denmark with regards to stage, metastasis, and mortality (see Supplementary Table II available online at http://informahealthcare.com/doi/abs/ 10.3109/0284186X.2014.973613, and Figure 1). Table II shows the odds ratio of cervical cancer according to fi laggrin genotype. FLG mutations were not associated with cervical cancer, neither for heterozygous nor the few homozygous carriers. The overall risk estimate was slightly reduced (crude OR 0.81, 95% CI 0.57 -1.12). Adjusting for age at blood sampling did not change the estimate substantially (OR 0.73, 95% CI 0.51 -1.05). As most cases had blood sampled after diagnosis, we obtained information on mortality after diagnosis, and weighted the OR for the genotype-specifi c inverse probability of survival between date of diagnosis and date of blood sampling. This method yielded an estimate close to unity (age-adjusted and weighted OR 0.96, 95% CI 0.58 -1.57).

Risk of cervical cancer
We examined possible reasons for the observed lack of association between FLG mutations and cervical cancer. However, neither residual confounding by age due to extremes of sampling age nor the different genotyping methods used appeared to infl uence results (Supplementary Table III

Prognosis of cervical cancer
As FLG mutations appeared to be associated with survival between diagnosis and sampling (i.e. see weighting of ORs in the fi rst paragraph of last section), we examined an association with cause-specifi c case mortality (Table III). We observed that the rate of death caused by cervical cancer was higher among FLG mutation carriers when compared to wildtype carriers (HR 4.55,, whereas this was not the case for other causes of death (HR 1.33, 95% CI 0. 27 -6.57). However, the association was attenuated after stratifi cation by cancer stage (HR 2.53,, and was based on fi ve deaths from cervical cancer and two from other causes among FLG mutation carriers (Table III). Figure 1 shows survival time since cervical cancer diagnosis, and illustrates that the former fi ve deaths took place in the fi rst year after diagnosis, and that death among other cases were comparable with death among all women with cervical cancer in Denmark.

Discussion
In this large case-control study, carriage of FLG mutations, and thus a possible impaired cervical mucosa barrier function, was not associated with an increased risk of cervical cancer as hypothesized. Our fi nding of a worse prognosis in cervical cancer patients carrying FLG mutations is interesting but should be interpreted with caution due to the possibility of an association with both severe cervical cancer and a worse prognosis. We suspected that a fi laggrin-reduced cervix increases HPV-access and -persistence and thereby the risk of cervical cancer. Although this hypothesis is not supported by the primary observation of a lack of association of FLG mutations with risk of cervical cancer, we believe the association with a worse prognosis lend support to the hypothesis, and is of interest because HPV-persistence has an important role in the pathogenesis of cervical cancer [35 -40]. Specifi cally, we observed that cases with FLG mutations had a four-fold higher rate of dying from cervical cancer than cases without FLG mutations. One explanation could be that mutation carriers more often develop severe cervical cancer, which in turn increases the risk of death. Another explanation could be that mutation carriers are poor responders to treatment, e.g. radiation therapy or chemotherapy. Our results suggest the former explanation is more likely, because the four-fold increase in the rate of death from cervical cancer attenuated after stratifi cation for cancer stage and severe cancer stages more often comprised mutation carriers. Both of these fi ndings were insignifi cant, showing that the statistical power for these extra analyses was limited. We recently observed that FLG mutations were associated with increased risk of all HPV-related pre-cancer In conclusion, carriage of common FLG mutations was not associated with the development of cervical cancer.

Declaration of interest:
This work was supported by the Danish Cancer Society (grant number R72-A4617-13-S2). PB is supported by an OAK Foundation Fellowship (grant number OCAY-12-319). JPT is supported by a Lundbeck Foundation Fellowship. and cancer among the present control populations including men, albeit with too few cervical cancer cases [41]. Overall, we believe these fi ndings should be interpreted with caution.
The large number of cases allowed us to study several cancer characteristics in detail using information from the Danish Cancer Registry. Among others we examined whether cases had another cancer before cervical cancer. If cancer initially did not start in the cervix, e.g. in the lungs where fi laggrin is not expressed, then we would not expect an association with FLG mutations. Although most cases did not have a previous cancer, results were the same when excluding those who had. We also studied the morphological type of cancer, and suspected FLG mutations would not be associated with cervical cancer that develops in the endocervical epithelia (i.e. adenocarcinomas) because fi laggrin is not normally expressed there but only in the exocervical epithelia where squamous cell carcinomas develop [7]. Regardless of cancer morphology, however, results were the same.
A characteristic of cases was the identifi cation in a hospital biobank storing excess from blood typings and antibody screenings; cases were therefore likely in for an operation. It might be argued that this could infl uence the generalizability of our fi ndings. However, our comparison with all cervix cancer patients in Denmark suggested that cases were representative in terms of metastasis and had only minor differences in cancer stage distributions (see Supplementary  Table II, available online at http://informahealthcare. com/doi/abs/10.3109/0284186X.2014.973613). Interestingly, our fi ndings were not materially different when including all the 778 cervical cancer patients we initially genotyped (to gain the largest possible statistical power notwithstanding case verifi cation). Finally, cases and controls were comparable for important epidemiological factors, i.e. blood sampling age and sex, and although they were from different birth years the frequency of FLG mutation carriers are comparable across birth years [42].
Overall, the study had several advantages, including its large size and statistical power, the use of cancer diagnosis verifi ed in the Danish Cancer Registry, the use of population-based controls, and the comparability with cervical cancer in Denmark for cancer stage, metastasis, and mortality after diagnosis (see Supplementary Table II available online at http:// informahealthcare.com/doi/abs/10.3109/0284186X. 2014.973613 and Figure 1). In addition, the study also to some extent reproduced known associations with FLG mutations, e.g. with atopic dermatitis (see Supplementary Table V, available online at http:// informahealthcare.com/doi/abs/10.3109/0284186X. 2014.973613) [2,43].