Advertisement

Combining CDKN1A gene expression and genome-wide SNPs in a twin cohort to gain insight into the heritability of individual radiosensitivity

  • Joanna Zyla
  • Sylwia Kabacik
  • Grainne O’Brien
  • Salma Wakil
  • Najla Al-Harbi
  • Jaakko Kaprio
  • Christophe Badie
  • Joanna PolanskaEmail author
  • Ghazi Alsbeih
Open Access
Original Article

Abstract

Individual variability in response to radiation exposure is recognised and has often been reported as important in treatment planning. Despite many efforts to identify biomarkers allowing the identification of radiation sensitive patients, it is not yet possible to distinguish them with certainty before the beginning of the radiotherapy treatment. A comprehensive analysis of genome-wide single-nucleotide polymorphisms (SNPs) and a transcriptional response to ionising radiation exposure in twins have the potential to identify such an individual. In the present work, we investigated SNP profile and CDKN1A gene expression in blood T lymphocytes from 130 healthy Caucasians with a complex level of individual kinship (unrelated, mono- or dizygotic twins). It was found that genetic variation accounts for 66% (95% CI 37–82%) of CDKN1A transcriptional response to radiation exposure. We developed a novel integrative multi-kinship strategy allowing investigating the role of genome-wide polymorphisms in transcriptomic radiation response, and it revealed that rs205543 (ETV6 gene), rs2287505 and rs1263612 (KLF7 gene) are significantly associated with CDKN1A expression level. The functional analysis revealed that rs6974232 (RPA3 gene), involved in mismatch repair (p value = 9.68e−04) as well as in RNA repair (p value = 1.4e−03) might have an important role in that process. Two missense polymorphisms with possible deleterious effect in humans were identified: rs1133833 (AKIP1 gene) and rs17362588 (CCDC141 gene). In summary, the data presented here support the validity of this novel integrative data analysis strategy to provide insights into the identification of SNPs potentially influencing radiation sensitivity. Further investigations in radiation response research at the genomic level should be therefore continued to confirm these findings.

Keywords

Radiation response CDKN1A p value integration Twin study GWAS 

Abbreviations

IR

ionising radiation

ROS

reactive oxygen species

DSBs

double-strand breaks

SNP

single-nucleotide polymorphism

QTL

quantitative trait loci

GWAS

genome-wide association study

unR

unRelated

DZ

DiZygotic twins

MZ

MonoZygotic twins

MAF

minor allele frequency

SEM

structural equation modelling

BIC

the Bayesian information criterion

LRT

likelihood ratio test

IBM

identical by model

nIBM

non-identical by model

TF

transcription factor

CI

confidence interval

GO

gene ontology

BP

biological process

KEGG

Kyoto Encyclopedia of Genes and Genomes

Introduction

Radiation therapy is a leading modality for cancer treatment. Although continuous technological improvements result in amelioration of radiotherapy protocols leading to precise tumour localisation and better dose delivery accuracy, patient inter-individual response to ionising radiation (IR) exposure is still a considerable risk factor (Pajic et al. 2015). Most patients do not present early, or late, normal tissue toxicity following radiotherapy and they are considered to be radioresistant. But a minority of patients develop severe complications during the course or at the end of the treatment, like skin erythema, nausea, diarrhoea and many others, after receiving a relatively low cumulative dose of radiation (Badie et al. 1995b; Lobachevsky et al. 2016). They are classified as radiosensitive. High-energy X-rays delivered to the cells cause water radiolysis and thereby production of reactive oxygen species (ROS) which indirectly damage DNA (Mettler 2012). The direct interaction between radiation and DNA leads to a range of DNA damage. Amongst them, double-strand breaks (DSBs) are the most toxic to the cells, leading to cell death or permanent cell cycle arrest if unrepaired. Therefore, efforts should be made to improve knowledge and identification of individuals sensitive to ionising radiation to improve radiation therapy efficiency and radiation protection (West and Barnett 2011). Individual radiosensitivity can be influenced by many factors such as DNA damage signalling and DNA repair (Vignard et al. 2013; Badie et al. 1995a, 1997; Morgan and Lawrence 2015), epigenetic modifications (Antwih et al. 2013) or genomic sequence variation (Curwen et al. 2010; Finnon et al. 2008). Some genes, mostly participating in DNA double-strand break repair process, were identified to be involved in human radiosensitivity, e.g. ATM, LIG4 and PRKDC (West and Barnett 2011). In this study, we focus on the expression CDKN1A (cyclin-dependent kinase inhibitor-1A) which encodes p21 protein and is regulated by p53 protein involved in cell cycle regulation and arrest following DNA damage (Cazzalini et al. 2010; Chen et al. 2015a; Galluzzi et al. 2016). CDKN1A also plays a crucial role in various cancer development (Abbas and Dutta 2009; Dunlop et al. 2012; Soltani et al. 2017). Several studies show an association between CDKN1A-SNPs and cancer and patient survival prognostics (e.g. Cazier et al. 2014; Kang et al. 2015; Vargas-Torres et al. 2016). A recent study of Price et al. (2015) suggests that CDKN1A regulates Langerhans cell and could influence the response of cutaneous tumours to radiotherapy. CDKN1A abnormal expression has been reported to be associated with acute sensitivity to radiation (Amundson et al. 2003; Badie et al. 2008; Szołtysek et al. 2018). In Alsbeih et al. (2007), they show that individual response in CDKN1A is related to inherent radiosensitivity. It is, therefore, assumed that CDKN1A expression level might be predictive of radiation toxicity and an investigation that allows explaining inter-patient CDKN1A expression variability is of high importance.

Many high-throughput approaches are currently used to gain an understanding of radiosensitivity; amongst them, the analysis of single-nucleotide polymorphisms (SNPs) is one of the most promising to investigate radiation response (Andreassen et al. 2012). Radiogenomics, which concentrates on the relation between genomics and radiation toxicity, has gained a high interest lately (West and Barnett 2011). Although a large number of studies have been reported (e.g. Best et al. 2011; Kerns et al. 2018; Mumbrekar et al. 2016; Rosenstein 2011), there is a need to continue identifying genes and SNPs that affect radiosensitivity to understand better the mechanism underlying radiation toxicity in sensitive patients. The choice of methods for data analysis allowing identification of relevant SNPs depends on the study design. Different statistical approaches have been widely discussed and presented (Bush and Moore 2012; Evangelou and Ioannidis 2013). Twin-based study designs were pointed as a promising source of information in genomics (Andrew et al. 2011; Bataille et al. 2012; Chen et al. 2015b; Tan et al. 2010) and transcriptomics (Majewska et al. 2017; Mamrut et al. 2017). In the following study, a dataset of a complex structure and small sample size with related (dizygotic and monozygotic twins) and unrelated individuals and quantitative measurement of CDKN1A gene expression as a metric of radio-toxicity is analysed. Such data structure is rarely studied and requires the development of dedicated signal analysis pipeline supporting the potential identification of a genetic signature of radiosensitivity. A literature screen revealed that a variety of quantitative trait loci (QTL) sib-pairs type methods are proposed to study related individuals (Kruglyak and Lander 1995a; Sham et al. 2002; Visscher and Hopper 2001). Several statistical approaches dedicated to the sample analysis of unrelated individuals are also available. We concluded that there is a lack of simple solutions available which would apply to complex study designs.

To fill that gap, we propose a novel signal analysis pipeline combining classical biometrical models (Kruglyak and Lander 1995b) and cross-sample p value integration methods. Although challenging, the integration approach appears to be the most promising methods in genome-wide studies (Moore et al. 2010; Stranger et al. 2011). The origin of integration methods arose from meta-analyses, where meta-genome-wide association studies (GWAS) brought new light to specific diseases (Barrett et al. 2009; Pharoah et al. 2013). Statistical integration in GWAS and SNP identification was previously presented as one of the most promising ways of analysis (Chen 2013; Chen et al. 2014; Zaykin and Kozbur 2010). In this study, we proposed to use statistical integration across individuals of different kinship for the validation of SNPs associated with radiation response. We demonstrated that the proposed procedure of integration improved the statistical analysis, especially in the case of small sample size studies. Finally, new promising candidate polymorphisms describing the association between genomics and radiation response in healthy individuals were identified.

Material and methods

Material

T lymphocytes were previously collected from healthy young adults of European ancestry sampled from the Finnish Twin Cohort Study (Finnon et al. 2008). The group under investigation here included 130 individuals divided into three subgroups according to their kinship: (1) 44 unrelated individuals (unR); (2) 28 dizygotic twin pairs (DZ) and (3) 15 monozygotic twin pairs (MZ). CDKN1A gene expression was measured for every individual by qPCR technique at two conditions: control (no irradiation (0 Gy)) and 2 h after sample irradiation with a single dose of 2 Gy of X-ray. The irradiation was performed at room temperature with an A.G.O. HS X-ray system by Aldermaston, Reading, UK—output 13 mA, 250 kV peak, 0.5 Gy/min. Detailed information about sample collection, storage and experiment was presented in (Kabacik et al. 2011; Manning et al. 2013). Additionally, DNA was extracted from all control samples using the DNeasy kit (Qiagen) and sent for genotyping. Analysis of 567,096 SNPs was performed by Axiom GW Human hg36.1 arrays (Affymetrix, ThermoFisher Scientific) according to manufacturer’ instruction. The used arrays did not include polymorphisms present in CDKN1A gene; thus, only SNPs in genes that interact with CDKN1A could be investigated in presented work.

Methods

Data pre-processing

All genotyped SNPs were annotated to the genome version 38 (according to NCBI resources). The standard GWAS specific quality control was performed, including minor allele frequency (MAF) control with level 10% and call rate on 90% (Turner et al. 2011). The quality control procedures reduced the number of SNPs from 567,096 to 383,322 (none of them was located in CDKN1A). The internally standardised ratio between the response at 2 Gy and referenced 0 Gy was calculated for investigated biomarker (CDKN1A) per each person. The 2 Gy vs 0 Gy ratio value will represent the radiation response of the investigated biomarker.

Heritability

At first, the hypothesis of the mean equality between MZ and DZ twin signals of 2 Gy vs 0 Gy ratio of CDKN1A expression was tested by a modified t test procedure proposed by Christian (1979). Further, the homogeneity of the MZ and DZ intra-class Pearson correlations was tested with the use of z-transformation (Fisher 1992). The assessment of genetic heritability of the trait was done by structural equation modelling (SEM) for the variance decomposition method, which bases on standard Falconer’s formula (Falconer 1965; Neale and Cardon 1994). The standard weights for additive (A) and dominant (D) genetic effects were set for monozygotic twins and equalled one for both effects. The 0.5 for additive effect and 0.25 dominant effect were considered for dizygotic twins. Common environment (C) weight values equal to 1 for both DZ and MZ twins as analysed twin pairs were reared together. The ACE and ADE models and all their submodels were constructed with the use of OpenMx (Neale et al. 2016). The Bayesian information criterion (BIC) was applied for model selection (Schwarz 1978). Additionally, the ADE and AE models were tested by a log-likelihood ratio test (LRT) for their over performance of the simple E model. To each model component, its 95% confidence interval (CI) was calculated.

Statistical analysis: unrelated

To verify the hypothesis on equality of signal means across observed genotypes, the adequate statistical test was performed on the probe of unrelated individuals (Bush and Moore 2012). The three different models of SNP-CDKN1A expression interactions were checked: genotype, dominant and recessive (Lettre et al. 2007; Zyla et al. 2014). Normality of CDKN1A expression’s distribution was calculated by the Shapiro-Wilk test, and homogeneity of variances was verified by Bartlett’s test or F test. Depending on their results, parametric (ANOVA, t test, the Welch test) or non-parametric (the Kruskal-Wallis, Mann-Whitney-Wilcoxon) tests were applied. The best model of SNP-CDKN1A interaction was assigned to each SNP based on calculated p values with the use of minimum p value criterion.

Statistical analysis: twin analysis

The novelty of presented work is stated for twin analysis. The SNP specific best model of SNP-CDKN1A interaction, obtained in the group of unrelated individuals for particular SNP, was used to split twin pairs into two subgroups named as identical by model (IBM) and non-identical by model (nIBM) following rules presented in Table 1. Splitting was done independently for each SNP; the difference of signal level (2 Gy vs 0 Gy ratio) between twins was calculated. In the case of IBM twins, the hypothesis on the average difference of signal between twins being equal to zero was verified. For nIBM twins, the null and alternative hypotheses depended on signal trend observed amongst unrelated individuals. For example, in the case of a significantly higher level of CDKN1A gene expression observed in Ax group vs. BB group in unrelated (unR) population, the same relation was tested in DZ and MZ subgroups by properly formulated one-side tests. The above-described procedure allows for response trend control in the process of signal validation. During the next step, the integration of p values from unR and DZ nIBM was performed. In the case of a dominant or recessive model of SNP-CDKN1A interaction, weighted z-method (Lipták 1958; Mosteller et al. 1954) with an inverse of standard error (1/SE) as the weighting factor was used (Whitlock 2005), while for genotype model, the Lancaster integration procedure was applied (Lancaster 1961). The procedure was not applied to data on monozygotic twins, who have the same genotype; hence, only identical by model twins were observed. Polymorphism was considered as associated with CDKN1A expression if unR and nIBM DZ combined p value was less than 0.001 and there was no evidence to reject the null hypothesis on equality of response between DZ and MZ at significance level α equal to 0.001. The diagram of the proposed analysis is presented in Fig. 1b. Finally, the results of integrative procedure were compared to commonly used non-parametric QTL method proposed by Kruglyak and Lander in (Kruglyak and Lander 1995b) including model weights presented in (Kruglyak and Lander 1995a). The Kruglyak and Lander method is the most common approach used in twin and sib-pair analysis, which allow to include models of genetic interactions. The candidate polymorphisms in this approach are selected as follows: p value unR less than 0.001, p value DZ QTL less than 0.001, p value MZ QTL less than 0.001. The diagram of the standard analysis is presented in Fig. 1a.
Table 1

The rules of splitting DZ and MZ twins into identical by model (IBM) and non-identical by model (nIBM) subgroups based on the best model of SNP-CDKN1A interaction found in unrelated population (unR). Letters A and B code for the genotyping results, A stands for reference allele, while B for mutant one

Sibling 1

Sibling 2

The best model of interaction in unR population

Genotype

Dominant, AA vs xB

Recessive, Ax vs BB

AA

AA

IBM

IBM

IBM

AA

AB

nIBM

nIBM

IBM

AA

BB

nIBM

nIBM

nIBM

AB

AA

nIBM

nIBM

IBM

AB

AB

IBM

IBM

IBM

AB

BB

nIBM

IBM

nIBM

BB

AA

nIBM

nIBM

nIBM

BB

AB

nIBM

IBM

nIBM

BB

BB

IBM

IBM

IBM

Fig. 1

The statistical analysis pipelines, where a represents the standard statistical analysis and b represents the developed novel statistical analysis procedure. Both are dedicated to the testing association in complex study design

In silico genomic functional analysis

In silico functional analysis was performed for sets of candidate radiation response relevant SNPs for signal 2 Gy vs 0 Gy level of CDKN1A expression and each analysis approach (standard and novel). The genomic location of each candidate SNP was assessed, and the lists of SNPs linked genes were constructed. Using the resources of SIGNOR 2.0 database (Perfetto et al. 2015), the list of genes which directly interact to/with CDKN1A was constructed. Additionally, the list of transcription factors (TFs) of CDKN1A gene was obtained using TRRUST 2.0 database (Han et al. 2017). Both lists were compared with obtained candidate polymorphisms. Additionally, the overrepresentation analysis of GO terms (biological process only) and KEGG pathways was performed (Falcon and Gentleman 2006; Kanehisa et al. 2016). The deleterious impact to the human organism of each candidate missense SNP was accessed by the PredictSNP algorithm (Bendl et al. 2014). Finally, the literature research was performed using the PubMed resource.

Results

Heritability

First, the intra-class correlation coefficients were calculated for both twin types, and hypothesis on MZ twins’ correlation being smaller or equal to DZ twins’ correlation was tested (H0: rMZrDZ). Correlations between twins for 2 Gy vs 0 Gy ratio equals 0.26 (DZ) and 0.77 (MZ) respectively. Within the monozygotic twins, significantly larger correlation than within dizygotic twins is observed (p value = 0.0140). It shows significantly larger signal similarity with increased genetic relatedness. Additionally, the equality of means of CDKN1A gene expression between DZ and MZ twins was tested. The Christian procedure brings no evidence against the hypothesis on equality of signal mean values (p value = 0.3333). Both outcomes allow for further investigation of heritability. The correlation coefficient for MZ is twice larger than for DZ, which determines the ADE model (and its submodels) to be only considered. The ADE model and its submodels where constructed, and the BIC method was used for model selection. The AE model shows the lowest BIC value and estimates the narrow-sense heritability estimate CDKN1A radiation response as equal to 66% (95% CI 37–82%). Detailed results for the main model and submodels are presented in Table 2. All the above support the hypothesis that a large fraction of CDKN1A response expression variation is accounted for genetic factors, which is of great importance for further association study.
Table 2

Result of heritability investigation for CDKN1A expression in response to radiation of dose 2 Gy (2 Gy vs 0 Gy ratio)

Model

BIC

A [95% CI]

D [95% CI]

E [95% CI]

LRT p value model vs E model

ADE

240

51 [0–82]

15 [0–81]

34 [0–63]

0.0005

AE

235

66 [37–82]

34 [0–62]

0.0001

E

246

100 [100–100]

Polymorphism investigation

The results of the analysis for investigated experimental condition (2 Gy vs 0 Gy ratio) and methods (integrative approach and non-parametric QTL approach as reference) are presented in Table 3. As can be observed, the novel method detects 1804 SNPs, of which 849 are located in transcriptomic regions. Out of all SNPs detected by a novel approach, 81% were also detected via standard approach. Figure 2 panel A presents the exemplary polymorphism detected by the standard approach and not identified by the novel method (rs710652 in KCNMB4 gene). The first plot (left panel) presents mean CDKN1A expression and its 95% CI in the phenotype-genotype interaction model for unrelated individuals. Second plot (middle panel) presents mean value and its 95% CI for expression difference between twin pairs. Third plot (right panel) shows CDKN1A expression within DZ nIBM twin pairs and serves as a validation of response trend found in unrelated individuals (left panel). As can be observed, the standard procedure detects polymorphisms, which do not validate by the trend of signal expression observed in unrelated individuals—it can be classified as false discovery. The list of all detected polymorphisms with their genomic information is included in Supplementary Material 1. In the next paragraph, a consideration of the relationship of detected polymorphisms and investigated phenomena is demonstrated.
Table 3

The data analysis results (after MZ twin validation) for both methods and signal at 2 Gy vs 0 Gy ratio. The first column represents the standard approach (Stand.), while the second column represents a novel integrative approach (Int.)

2 Gy vs 0 Gy ratio

Genotype

Dominant

Recessive

Total

Common

Stand

Int

Stand

Int

Stand

Int

Stand

Int

Initially, # of SNPs

2093

177,481

203,748

383,322

α = 0.001

# candidate SNPs

1

52

92

839

88

913

181

1804

147 [81%]

# SNPs in genes

1

25

50

406

45

418

96

849

78 [81%]

# unique protein-coding genes

 

81

615

74 [91%]

Fig. 2

Levels of signal response (2 Gy vs 0 Gy) in the recessive genetic model under different genotypes and different kinship classes for a rs710652 polymorphism in KCNMB4, b rs205543 in ETV6, c rs1263612 in KLF7 and d rs6974232 in RPA3 genes. The two left-side plots represent the 95% confidence interval for the mean of CDKN1A gene expression. The right-side plots represent the expression levels for non-identical by model (nIBM) dizygotic twins, where discontinued green colour lines represent identical response trend while discontinued red colour lines represent opposite response trend amongst unR and DZ nIBM

In silico functional analysis

A literature study was performed to identify the signalling cascade from the genes with the candidate relevant SNPs to the CDKN1A gene. Two types of linkage were studied: (1) interaction, where control of CDKN1A expression is done by transcription factor (TFs) or protein phosphorylation with identified SNPs and (2) complex, where a group of genes (with detected SNPs) show an overrepresentation of pathway highly relevant to radiation toxicity. Finally, they are accompanied by missense polymorphism investigation. The list of proteins taking part in direct phosphorylation of p21 (encoded by CDKN1A) was obtained using the resources of SIGNOR 2.0 (Perfetto et al. 2015), while TF genes for CDKN1A were established via TRRUST 2.0 (Han et al. 2017). From the group of detected polymorphisms, three of them are located in genes responsible for transcription regulation of CDKN1A (Table 4). First SNP—rs205543—is located in the ETV6 gene also known as TEL oncogene. TEL oncogene was shown as TF of CDKN1A and BBC3 and is related to “transcriptional misregulation in cancer” pathway (Yamagata et al. 2006). The rearrangements of ETV6 were also observed in radiation-associated thyroid cancer (Leeman-Neill et al. 2014). Next, two SNPs (rs2287505 and rs1263612) are located in KLF7 gene (part of the Kruppel family), which is mainly responsible for cell proliferation, and it transcriptionally regulates CDKN1A expression (Smaldone et al. 2004). CDKN1A expression level in different kinship subgroups and genotypes for polymorphisms in ETV6 and KLF7 genes are presented in Fig. 2. As it can be observed in panels B and C for DZ nIBM twins (middle panel), CIs do not include zero value, which confirms different response caused by different genotypes at a significance level < 0.05. Moreover, the right-side panel plot shows that nIBM dizygotic twins represent the same signal trend as observed in unrelated individuals. For the differences of CDKN1A response signal in IBM twin pairs, there is no statistical evidence that it is different from zero. It fulfils the expectation as identical twins that share the same genotype model express the similar CDKN1A radiation response. None of the presented polymorphisms was detected by the standard approach.
Table 4

Result of the investigation on transcription factors and phosphorylation proteins

Gene

rs ID

Model of interaction

Type of interaction with CDKN1Aa

Integrated p value

Ref

ETV6

205543

AX vs BB

TF

4.39e−04

(Yamagata et al. 2006)

KLF7

2287505

AX vs BB

RoTL

8.56e−04

(Smaldone et al. 2004)

1263612

AX vs BB

9.27e−04

aTF transcription factor, RoTL Regulation on transcription level

Next, the overrepresentation analysis for all obtained genes with SNPs presented in Supplementary Material 1 was performed on KEGG and GO (biological process (BP) only) resources (Table 5). A detailed list is included in Supplementary Material 2. As can be observed, the novel integrative method shows a higher number of overrepresented pathways and GO terms when compared to the standard approach. Out of overrepresented pathways at 2 Gy vs 0 Gy ratio in KEGG and gene ontology (GO) those indicated by RPA3 gene (with candidate SNPs rs6974232) are highly related to the investigated phenomenon. RPA3 plays a role in both DNA replication and the cellular response to DNA damage (together with RPA1 and RPA2). In the cellular response to DNA damage, the RPA complex controls DNA repair and DNA damage checkpoint activation (Haring et al. 2008). Recently, Guo et al. showed the relationship between RPA family and distant metastasis in nasopharyngeal carcinoma patients treated with intensity-modulated radiation therapy (Guo et al. 2016). Of the overrepresented KEGG pathways with RPA3 involvement, we can distinguish mismatch repair (p value = 9.68e−04) or DNA replication (p value = 2.63e−02). In the case of GO analysis, the following terms were identified: RNA repair (p value = 1.4e−03), mismatch repair (p value = 1.52e−02) and nucleotide excision repair by DNA gap filling (p value = 3.25e−02). As mentioned in the introduction, the DNA repair processes and cell cycle control are crucial for radiosensitivity phenomenon. RPA3 occurs together with investigated CDKN1A in Reactome pathways (Fabregat et al. 2017): mitotic G1-G1/S phases, G1/S transition and cell cycle checkpoints. The response level of CDKN1A under different kinship and polymorphism rs6974232 is presented in Fig. 2 panel D.
Table 5

The summary results for overrepresentation analysis

 

KEGG

GO [BP]

Standard

4

99

Integrative

46

399

Common

4

17

Finally, the missense SNPs were investigated by PredictSNP to assess the possible deleterious impact on protein function. Out of 21 missense polymorphisms, the rs1133833, which change the arginine in position 23 to threonine (R23T) in AKIP1 gene, was predicted as deleterious with a score of 72%. The AKIP1 gene encodes A-kinase-interacting protein 1 which regulates the effect of the cAMP-dependent protein kinase signalling pathway on the NF-κB activation cascade. It is well known that IR activates the NF-κB pathway which further makes cancer cell resistant to treatment, while in parallel, the NF-κB has an impact to apoptosis control (Gao et al. 2010; Magné et al. 2006; Molavi Pordanjani and Jalal Hosseinimehr 2016). Additionally, the AKIP1 is overexpressed in breast cancer and is related to poor prognosis of survival (Mo et al. 2016). Second, a deleterious polymorphism was rs17362588 located in CCDC141 gene, and it changes arginine in position 935 to tryptophan (R935W; score 87%). The CCDC141 encodes a coiled-coil domain-containing protein. However, its role is as yet unclear. Several studies show mutations in CCDC141 in patients with thyroid disorder known as idiopathic hypogonadotropic hypogonadism (Hutchins et al. 2016; Turan et al. 2017). However, in relation to radiation response, apoptosis and CDKN1A have not been described in the literature.

Discussion and conclusions

The work presented here investigated genetic component in CDKN1A expression following ionising radiation exposure which was used as a surrogate marker for radiosensitivity of healthy individuals. Firstly, we have shown that CDKN1A transcriptional response to radiation is heritable, with a heritability estimate of 66% (95% CI 37−82%) based on a twin analysis. This provided motivation for further investigation at the genomic level (SNP investigation). Additionally, those findings are consistent with previous investigations of heritability for apoptosis and cell cycle delay (Camplejohn et al. 2006; Finnon et al. 2008) and brought new insight of understanding which genes can be responsible for previously observed outcomes. Furthermore, we proposed here a novel signal analysis pipeline for quantitative genomic association analysis of data with different kinship and no family information. The presented workflow is a combination of SNP genotype modelling and statistical integration. It can be an alternative for well-known linkage analysis of sib-pairs, when, in most of the cases, family information is required (Fulker et al. 1999; Li et al. 2005). Additionally, the integration process increases the power of the conducted analysis, which is of great importance when the sample size is small. Finally, the method proposed here includes control of response trends in the process of validation, which allows for reliable candidate polymorphism detection, reducing the number of false positives. The in silico investigation showed that obtained polymorphisms are related to the investigated phenomenon at the global scale via overrepresentation analysis of pathways and gene ontologies. Additionally, the direct interaction with analysed CDKN1A expression was shown. SNPs located in CDKN1A transcription factors genes, ETV6 (rs205543) and KLF7 (rs2287505, rs1263612), are of special interests for further biological investigation. Further, the rs6974232 in RPA3 gene should be highlighted as it participates in DNA repair and replication processes which are crucial pathways to radiation response. Finally, the missense polymorphism rs1133833 in AKIP1 gene with possible deleterious impact to protein function was identified. In summary, the results presented support the validity of the proposed statistical strategy of analysis and demonstrate that high-throughput genomic approaches, such as the one described here, can provide insights to identify radiosensitive patients, and further similar investigations will help to develop future predictive assays for clinical applications.

Notes

Author contributions

Conceived of designed study: Ghazi Alsbeih and Christophe Badie

Performed research: Sylwia Kabacik, Grainne O’Brien, Salma Wakil and Najla Al-Harbi

Analysed data: Joanna Polanska and Joanna Zyla

Contributed new methods or models: Jaakko Kaprio, Joanna Polanska and Joanna Zyla

Wrote the paper: Joanna Zyla, Joanna Polanska, Ghazi Alsbeih and Christophe Badie

Funding information

This work was funded by the National Science Centre, Poland grant 2013/08/M/ST6/00924 (JZ) and SUT grant 02/010/BK18/0102/8 (JP); the National Science, Technology and Innovation Plan (NSTIP), grant 11-BIO1429-20 and RAC# 2120 003 (SW, GA); and the Academy of Finland grant 308248, 312073 (JK). Calculations were carried out using the infrastructure of GeCONiI (POIG.02.03.01-24-099/13).

Compliance with ethical standards

Competing interests

The authors declare that they have no competing interests.

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Supplementary material

10142_2019_658_MOESM1_ESM.xls (614 kb)
ESM 1 (XLS 614 kb)
10142_2019_658_MOESM2_ESM.xls (80 kb)
ESM 2 (XLS 80 kb)

References

  1. Abbas T, Dutta A (2009) p21 in cancer: intricate networks and multiple activities. Nat Rev Cancer 9:400–414Google Scholar
  2. Alsbeih G, Torres M, Al-Harbi N, Al-Buhairi M (2007) Evidence that individual variations in TP53 and CDKN1A protein responsiveness are related to inherent radiation sensitivity. Radiat Res 167:58–65Google Scholar
  3. Amundson SA, Lee RA, Koch-Paiz CA, Bittner ML, Meltzer P, Trent JM, Fornace AJ (2003) Differential responses of stress genes to low dose-rate γ irradiation1 1 DOE grant ER62683. Mol Cancer Res 1:445–452Google Scholar
  4. Andreassen CN, Dikomey E, Parliament M, West CML (2012) Will SNPs be useful predictors of normal tissue radiosensitivity in the future? Radiother Oncol 105:283–288Google Scholar
  5. Andrew T, Calloway CD, Stuart S, Lee SH, Gill R, Clement G, Chowienczyk P, Spector TD, Valdes AM (2011) A twin study of mitochondrial DNA polymorphisms shows that heteroplasmy at multiple sites is associated with mtDNA variant 16,093 but not with zygosity. PLoS One 6:e22332Google Scholar
  6. Antwih DA, Gabbara KM, Lancaster WD, Ruden DM, Zielske SP (2013) Radiation-induced epigenetic DNA methylation modification of radiation-response pathways. Epigenetics 8:839–848Google Scholar
  7. Badie C, Iliakis G, Foray N, Alsbeih G, Cedervall B, Chavaudra N, Pantelias G, Arlett C, Malaise EP (1995a) Induction and rejoining of DNA double-strand breaks and interphase chromosome breaks after exposure to X rays in one normal and two hypersensitive human fibroblast cell lines. Radiat Res 144:26–35Google Scholar
  8. Badie C, Iliakis G, Foray N, Alsbeih G, Pantellias GE, Okayasu R, Cheong N, Russell NS, Begg AC, Arlett CF (1995b) Defective repair of DNA double-strand breaks and chromosome damage in fibroblasts from a radiosensitive leukemia patient. Cancer Res 55:1232–1234Google Scholar
  9. Badie C, Goodhardt M, Waugh A, Doyen N, Foray N, Calsou P, Singleton B, Gell D, Salles B, Jeggo P, Arlett CF, Malaise EP (1997) A DNA double-strand break defective fibroblast cell line (180BR) derived from a radiosensitive patient represents a new mutant phenotype. Cancer Res 57:4600–4607Google Scholar
  10. Badie C, Dziwura S, Raffy C, Tsigani T, Alsbeih G, Moody J, Finnon P, Levine E, Scott D, Bouffler S (2008) Aberrant CDKN1A transcriptional response associates with abnormal sensitivity to radiation treatment. Br J Cancer 98:1845–1851Google Scholar
  11. Barrett JC et al (2009) Genome-wide association study and meta-analysis find that over 40 loci affect risk of type 1 diabetes. Nat Genet 41:703–707Google Scholar
  12. Bataille V, Lens M, Spector T (2012) The use of the twin model to investigate the genetics and epigenetics of skin diseases with genomic, transcriptomic and methylation data. J Eur Acad Dermatol Venereol 26:1067–1073Google Scholar
  13. Bendl J, Stourac J, Salanda O, Pavelka A, Wieben ED, Zendulka J, Brezovsky J, Damborsky J (2014) PredictSNP: robust and accurate consensus classifier for prediction of disease-related mutations. PLoS Comput Biol 10:e1003440Google Scholar
  14. Best T, Li D, Skol AD, Kirchhoff T, Jackson SA, Yasui Y, Bhatia S, Strong LC, Domchek SM, Nathanson KL, Olopade OI, Huang RS, Mack TM, Conti DV, Offit K, Cozen W, Robison LL, Onel K (2011) Variants at 6q21 implicate PRDM1 in the etiology of therapy-induced second malignancies after Hodgkin’s lymphoma. Nat Med 17:941–943Google Scholar
  15. Bush WS, Moore JH (2012) Genome-wide association studies. PLoS Comput Biol 8:e1002822Google Scholar
  16. Camplejohn R, Hodgson S, Carter N, Kato B, Spector T (2006) Heritability of DNA-damage-induced apoptosis and its relationship with age in lymphocytes from female twins. Br J Cancer 95:520–524Google Scholar
  17. Cazier J-B et al (2014) Whole-genome sequencing of bladder cancers reveals somatic CDKN1A mutations and clinicopathological associations with mutation burden. Nat Commun 5:3756Google Scholar
  18. Cazzalini O, Scovassi AI, Savio M, Stivala LA, Prosperi E (2010) Multiple roles of the cell cycle inhibitor p21CDKN1A in the DNA damage response. Mutat Res Rev Mutat Res 704:12–20Google Scholar
  19. Chen Z (2013) Association tests through combining p-values for case control genome-wide association studies. Statist Probab Lett 83:1854–1862Google Scholar
  20. Chen Z, Yang W, Liu Q, Yang JY, Li J, Yang MQ (2014) A new statistical approach to combining p-values using gamma distribution and its application to genome-wide association study. BMC Bioinf 15:S3Google Scholar
  21. Chen A et al (2015a) The role of p21 in apoptosis, proliferation, cell cycle arrest, and antioxidant activity in UVB-irradiated human HaCaT keratinocytes. Med Sci Monit Basic Res 21:86–95Google Scholar
  22. Chen X, Kuja-Halkola R, Rahman I, Arpegård J, Viktorin A, Karlsson R, Hägg S, Svensson P, Pedersen NL, Magnusson PKE (2015b) Dominant genetic variation and missing heritability for human complex traits: insights from twin versus genome-wide common SNP models. Am J Hum Genet 97:708–714Google Scholar
  23. Christian JC (1979) Testing twin means and estimating genetic variance. Basic methodology for the analysis of quantitative twin data. Acta Genet Med Gemellol: twin research 28:35–40Google Scholar
  24. Curwen GB, Cadwell KK, Winther JF, Janet Tawn E, Rees GS, Olsen JH, Rechnitzer C, Schroeder H, Guldberg P, Cordell HJ, Boice JD Jr (2010) The heritability of G2 chromosomal radiosensitivity and its association with cancer in Danish cancer survivors and their offspring. Int J Radiat Biol 86:986–995Google Scholar
  25. Dunlop MG et al (2012) Common variation near CDKN1A, POLD3 and SHROOM2 influences colorectal cancer risk. Nat Genet 44:770–776Google Scholar
  26. Evangelou E, Ioannidis JP (2013) Meta-analysis methods for genome-wide association studies and beyond. Nat Rev Genet 14:379–389Google Scholar
  27. Fabregat A et al (2017) The reactome pathway knowledgebase. Nucleic Acids Res 46:D649–D655Google Scholar
  28. Falcon S, Gentleman R (2006) Using GOstats to test gene lists for GO term association. Bioinformatics 23:257–258Google Scholar
  29. Falconer DS (1965) The inheritance of liability to certain diseases, estimated from the incidence among relatives. Ann Hum Genet 29:51–76Google Scholar
  30. Finnon P, Robertson N, Dziwura S, Raffy C, Zhang W, Ainsbury L, Kaprio J, Badie C, Bouffler S (2008) Evidence for significant heritability of apoptotic and cell cycle responses to ionising radiation. Hum Genet 123:485–493Google Scholar
  31. Fisher RA (1992) Statistical methods for research workers. In: Breakthroughs in statistics. Springer, pp 66–70Google Scholar
  32. Fulker D, Cherny S, Sham P, Hewitt J (1999) Combined linkage and association sib-pair analysis for quantitative traits. Am J Hum Genet 64:259–267Google Scholar
  33. Galluzzi L, Bravo-San Pedro JM, Kepp O, Kroemer G (2016) Regulated cell death and adaptive stress responses. Cell Mol Life Sci 73:2405–2410Google Scholar
  34. Gao N, Hibi Y, Cueno ME, Asamitsu K, Okamoto T (2010) A-kinase interacting protein 1 (AKIP1) acts as a molecular determinant of the role of PKA in NF-κB signaling. J Biol Chem 285(36):28097–28104.  https://doi.org/10.1074/jbc.M110.116566 Google Scholar
  35. Guo Q, Lu T, Chen Y, Su Y, Zheng Y, Chen Z, Chen C, Lin S, Pan J, Yuan X (2016) Genetic variations in the PI3K-PTEN-AKT-mTOR pathway are associated with distant metastasis in nasopharyngeal carcinoma patients treated with intensity-modulated radiation therapy. Sci Rep 6:37576.  https://doi.org/10.1038/srep37576 Google Scholar
  36. Han H et al (2017) TRRUST v2: an expanded reference database of human and mouse transcriptional regulatory interactions. Nucleic Acids Res 46:D380–D386Google Scholar
  37. Haring SJ, Mason AC, Binz SK, Wold MS (2008) Cellular functions of human RPA1: multiple roles of domains in replication, repair, and checkpoints. J Biol Chem 283(27):19095–19111.  https://doi.org/10.1074/jbc.M800881200 Google Scholar
  38. Hutchins BI, Kotan LD, Taylor-Burds C, Ozkan Y, Cheng PJ, Gurbuz F, Tiong JDR, Mengen E, Yuksel B, Topaloglu AK, Wray S (2016) CCDC141 mutation identified in anosmic hypogonadotropic hypogonadism (Kallmann syndrome) alters GnRH neuronal migration. Endocrinology 157:1956–1966Google Scholar
  39. Kabacik S, Mackay A, Tamber N, Manning G, Finnon P, Paillier F, Ashworth A, Bouffler S, Badie C (2011) Gene expression following ionising radiation: identification of biomarkers for dose estimation and prediction of individual response. Int J Radiat Biol 87:115–129Google Scholar
  40. Kanehisa M, Furumichi M, Tanabe M, Sato Y, Morishima K (2016) KEGG: new perspectives on genomes, pathways, diseases and drugs. Nucleic Acids Res 45:D353–D361Google Scholar
  41. Kang BW, Jeon HS, Chae YS, Lee SJ, Park JY, Choi JE, Park JS, Choi GS, Kim JG (2015) Association between GWAS-identified genetic variations and disease prognosis for patients with colorectal cancer. PLoS One 10:e0119649Google Scholar
  42. Kerns S et al (2018) Radiation biology and oncology in the genomic era. Br J Radiol 91:20170949Google Scholar
  43. Kruglyak L, Lander ES (1995a) Complete multipoint sib-pair analysis of qualitative and quantitative traits. Am J Hum Genet 57:439Google Scholar
  44. Kruglyak L, Lander ES (1995b) A nonparametric approach for mapping quantitative trait loci. Genetics 139:1421–1428Google Scholar
  45. Lancaster H (1961) The combination of probabilities: an application of orthonormal functions. Austral J Statist 3:20–33Google Scholar
  46. Leeman-Neill RJ, Kelly LM, Liu P, Brenner AV, Little MP, Bogdanova TI, Evdokimova VN, Hatch M, Zurnadzy LY, Nikiforova MN, Yue NJ, Zhang M, Mabuchi K, Tronko MD, Nikiforov YE (2014) ETV6-NTRK3 is a common chromosomal rearrangement in radiation-associated thyroid cancer. Cancer 120:799–807Google Scholar
  47. Lettre G, Lange C, Hirschhorn JN (2007) Genetic model testing and statistical power in population-based association studies of quantitative traits Genetic Epidemiology: The Official Publication of the International Genetic Epidemiology. Society 31:358–362Google Scholar
  48. Li M, Boehnke M, Abecasis GR (2005) Joint modeling of linkage and association: identifying SNPs responsible for a linkage signal. Am J Hum Genet 76:934–949Google Scholar
  49. Lipták T (1958) On the combination of independent tests. Magyar Tud Akad Nat Kutato int Kozl 3:171–197Google Scholar
  50. Lobachevsky P, Leong T, Daly P, Smith J, Best N, Tomaszewski J, Thompson ER, Li N, Campbell IG, Martin RF, Martin OA (2016) Compromized DNA repair as a basis for identification of cancer radiotherapy patients with extreme radiosensitivity. Cancer Lett 383:212–219Google Scholar
  51. Magné N, Toillon R-A, Bottero V, Didelot C, Van Houtte P, Gérard J-P, Peyron J-F (2006) NF-κB modulation and ionizing radiation: mechanisms and future directions for cancer treatment. Cancer Lett 231:158–168Google Scholar
  52. Majewska M, Lipka A, Paukszto L, Jastrzebski JP, Myszczynski K, Gowkielewicz M, Jozwik M, Majewski MK (2017) Transcriptome profile of the human placenta. Funct Integr Genomics 17:551–563Google Scholar
  53. Mamrut S, Avidan N, Truffault F, Staun-Ram E, Sharshar T, Eymard B, Frenkian M, Pitha J, de Baets M, Servais L, Berrih-Aknin S, Miller A (2017) Methylome and transcriptome profiling in myasthenia gravis monozygotic twins. J Autoimmun 82:62–73Google Scholar
  54. Manning G, Kabacik S, Finnon P, Bouffler S, Badie C (2013) High and low dose responses of transcriptional biomarkers in ex vivo X-irradiated human blood. Int J Radiat Biol 89:512–522Google Scholar
  55. Mettler FA (2012) Medical effects and risks of exposure to ionising radiation. J Radiol Prot 32:N9–N13Google Scholar
  56. Mo D et al (2016) Overexpression of AKIP1 predicts poor prognosis of patients with breast carcinoma and promotes cancer metastasis through Akt/GSK-3β/Snail pathway. Am J Transl Res 8:4951Google Scholar
  57. Molavi Pordanjani S, Jalal Hosseinimehr S (2016) The role of NF-κB inhibitors in cell response to radiation. Curr Med Chem 23:3951–3963Google Scholar
  58. Moore JH, Asselbergs FW, Williams SM (2010) Bioinformatics challenges for genome-wide association studies. Bioinformatics 26:445–455Google Scholar
  59. Morgan MA, Lawrence TS (2015) Molecular pathways: overcoming radiation resistance by targeting DNA damage response pathways. Clin Cancer Res 21:2898–2904Google Scholar
  60. Mosteller F, Green BF, Bush RR (1954) Selected quantitative techniques: by Frederick Mosteller and Robert R. Bush-and attitude measurement By Bert F. Green. Addison-WesleyGoogle Scholar
  61. Mumbrekar KD, Goutham HV, Vadhiraja BM, Sadashiva SRB (2016) Polymorphisms in double strand break repair related genes influence radiosensitivity phenotype in lymphocytes from healthy individuals. DNA Repair 40:27–34Google Scholar
  62. Neale MC, Cardon LR (1994) Methodology for genetic studies of twins and families. Stat Med 13:199–199Google Scholar
  63. Neale MC, Hunter MD, Pritikin JN, Zahery M, Brick TR, Kirkpatrick RM, Estabrook R, Bates TC, Maes HH, Boker SM (2016) OpenMx 2.0: Extended structural equation and statistical modeling. Psychometrika 81:535–549Google Scholar
  64. Pajic J, Rakic B, Rovcanin B, Jovicic D, Novakovic I, Milovanovic A, Pajic V (2015) Inter-individual variability in the response of human peripheral blood lymphocytes to ionizing radiation: comparison of the dicentric and micronucleus assays. Radiat Environ Biophys 54:317–325Google Scholar
  65. Perfetto L et al (2015) SIGNOR: a database of causal relationships between biological entities. Nucleic Acids Res 44:D548–D554Google Scholar
  66. Pharoah PD et al (2013) GWAS meta-analysis and replication identifies three new susceptibility loci for ovarian cancer. Nat Genet 45:362–370Google Scholar
  67. Price JG, Idoyaga J, Salmon H, Hogstad B, Bigarella CL, Ghaffari S, Leboeuf M, Merad M (2015) CDKN1A regulates Langerhans cell survival and promotes T reg cell generation upon exposure to ionizing irradiation. Nat Immunol 16:1060–1068Google Scholar
  68. Rosenstein BS (2011) Identification of SNPs associated with susceptibility for development of adverse reactions to radiotherapy. Pharmacogenomics 12:267–275Google Scholar
  69. Schwarz G (1978) Estimating the dimension of a model. Ann Stat 6:461–464Google Scholar
  70. Sham PC, Purcell S, Cherny SS, Abecasis GR (2002) Powerful regression-based quantitative-trait linkage analysis of general pedigrees. Am J Hum Genet 71:238–253Google Scholar
  71. Smaldone S, Laub F, Else C, Dragomir C, Ramirez F (2004) Identification of MoKA, a novel F-box protein that modulates Krüppel-like transcription factor 7 activity. Mol Cell Biol 24:1058–1069Google Scholar
  72. Soltani I, Gharbi H, Hassine IB, Bouguerra G, Douzi K, Teber M, Abbes S, Menif S (2017) Regulatory network analysis of microRNAs and genes in imatinib-resistant chronic myeloid leukemia. Funct Integr Genomics 17:263–277Google Scholar
  73. Stranger BE, Stahl EA, Raj T. (2011) Progress and promise of genome-wide association studies for human complex trait genetics. Genetics 187(2):367–383Google Scholar
  74. Szołtysek K, Janus P, Zając G, Stokowy T, Walaszczyk A, Widłak W, Wojtaś B, Gielniewski B, Cockell S, Perkins ND, Kimmel M, Widlak P (2018) RRAD, IL4I1, CDKN1A, and SERPINE1 genes are potentially co-regulated by NF-κB and p53 transcription factors in cells exposed to high doses of ionizing radiation. BMC Genomics 19:813Google Scholar
  75. Tan Q, Kyvik KO, Kruse TA, Christensen K (2010) Dissecting complex phenotypes using the genomics of twins. Funct Integr Genomics 10:321–327Google Scholar
  76. Turan I, Hutchins BI, Hacihamdioglu B, Kotan LD, Gurbuz F, Ulubay A, Mengen E, Yuksel B, Wray S, Topaloglu AK (2017) CCDC141 mutations in idiopathic hypogonadotropic hypogonadism. J Clin Endocrinol Metab 102:1816–1825Google Scholar
  77. Turner S, Armstrong LL, Bradford Y, Carlson CS, Crawford DC, Crenshaw AT, de Andrade M, Doheny KF, Haines JL, Hayes G, Jarvik G, Jiang L, Kullo IJ, Li R, Ling H, Manolio TA, Matsumoto M, McCarty CA, McDavid AN, Mirel DB, Paschall JE, Pugh EW, Rasmussen LV, Wilke RA, Zuvich RL, Ritchie MD (2011) Quality control procedures for genome-wide association studies. Curr Protoc Hum Genet 68(1):1.19.1–1.19.18Google Scholar
  78. Vargas-Torres SL, Portari EA, Silva AL, Klumb EM, da Rocha Guillobel HC, de Camargo MJ, Santos-Rebouças CB, Russomano FB, Macedo JMB (2016) Roles of CDKN1A gene polymorphisms (rs1801270 and rs1059234) in the development of cervical neoplasia. Tumor Biol 37:10469–10478Google Scholar
  79. Vignard J, Mirey G, Salles B (2013) Ionizing-radiation induced DNA double-strand breaks: a direct and indirect lighting up. Radiother Oncol 108:362–369Google Scholar
  80. Visscher P, Hopper J (2001) Power of regression and maximum likelihood methods to map QTL from sib-pair and DZ twin data. Ann Hum Genet 65:583–601Google Scholar
  81. West CM, Barnett GC (2011) Genetics and genomics of radiotherapy toxicity: towards prediction. Genome Med 3:52Google Scholar
  82. Whitlock MC (2005) Combining probability from independent tests: the weighted Z-method is superior to Fisher’s approach. J Evol Biol 18:1368–1373Google Scholar
  83. Yamagata T, Maki K, Waga K, Mitani K (2006) TEL/ETV6 induces apoptosis in 32D cells through p53-dependent pathways. Biochem Biophys Res Commun 347:517–526Google Scholar
  84. Zaykin DV, Kozbur DO (2010) P-value based analysis for shared controls design in genome-wide association studies. Genet Epidemiol 34:725–738Google Scholar
  85. Zyla J, Badie C, Alsbeih G, Polanska J (2014) Modelling of genetic interactions in GWAS reveals more complex relations between genotype and phenotype. In: Proceedings of the International Joint Conference on Biomedical Engineering Systems and Technologies, vol 3. SCITEPRESS-Science and Technology Publications, Lda, pp 204–208Google Scholar

Copyright information

© Springer-Verlag GmbH Germany, part of Springer Nature 2019

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Authors and Affiliations

  1. 1.Data Mining Division, Faculty of Automatic Control, Electronic and Computer ScienceSilesian University of TechnologyGliwicePoland
  2. 2.Cellular Biology Group, Radiation Effects Department, Centre for Radiation, Chemical and Environmental HazardsPublic Health EnglandDidcotUK
  3. 3.Cancer Mechanisms and Biomarkers Group, Radiation Effects Department, Centre for Radiation, Chemical and Environmental HazardsPublic Health EnglandDidcotUK
  4. 4.Department of GeneticsKing Faisal Specialist Hospital and Research CentreRiyadhKingdom of Saudi Arabia
  5. 5.Radiation Biology Section, Biomedical Physics DepartmentKing Faisal Specialist Hospital and Research CentreRiyadhKingdom of Saudi Arabia
  6. 6.Department of Public Health and Institute for Molecular Medicine FIMMUniversity of HelsinkiHelsinkiFinland

Personalised recommendations