Micronuclei Formation upon Radioiodine Therapy for Well-Differentiated Thyroid Cancer: The Influence of DNA Repair Genes Variants

Radioiodine therapy with 131I remains the mainstay of standard treatment for well-differentiated thyroid cancer (DTC). Prognosis is good but concern exists that 131I-emitted ionizing radiation may induce double-strand breaks in extra-thyroidal tissues, increasing the risk of secondary malignancies. We, therefore, sought to evaluate the induction and 2-year persistence of micronuclei (MN) in lymphocytes from 26 131I-treated DTC patients and the potential impact of nine homologous recombination (HR), non-homologous end-joining (NHEJ), and mismatch repair (MMR) polymorphisms on MN levels. MN frequency was determined by the cytokinesis-blocked micronucleus assay while genotyping was performed through pre-designed TaqMan® Assays or conventional PCR-restriction fragment length polymorphism (RFLP). MN levels increased significantly one month after therapy and remained persistently higher than baseline for 2 years. A marked reduction in lymphocyte proliferation capacity was also apparent 2 years after therapy. MLH1 rs1799977 was associated with MN frequency (absolute or net variation) one month after therapy, in two independent groups. Significant associations were also observed for MSH3 rs26279, MSH4 rs5745325, NBN rs1805794, and tumor histotype. Overall, our results suggest that 131I therapy may pose a long-term challenge to cells other than thyrocytes and that the individual genetic profile may influence 131I sensitivity, hence its risk-benefit ratio. Further studies are warranted to confirm the potential utility of these single nucleotide polymorphisms (SNPs) as radiogenomic biomarkers in the personalization of radioiodine therapy.


Introduction
Thyroid cancer (TC) is the most common endocrine malignancy, accounting for approximately 2.1% of cancers diagnosed all over the world. TC incidence is about two to four times higher in women

Study Population
Twenty-six DTC patients proposed for radioiodine therapy at the Department of Nuclear Medicine of the Portuguese Oncology Institute of Lisbon (Portugal) were selected according to criteria published elsewhere [22]. All participants were treated according to current practice, consisting of total thyroidectomy followed by oral administration of 131 I, 70 mCi (15 patients) or 100 mCi (11 patients), to ablate thyroid remnant cells. Patients were followed for two years unless they had to be submitted to further treatment. In such cases, patients were no longer elective for cytogenetic analysis and had to be excluded from further analysis. A mixed cross-sectional and longitudinal study design was used, respectively, for comparisons among genotypes or dose groups at each time point and across different time points. In the latter case, pre-treatment values allowed each patient to serve as his own control.
To characterize the study population and account for potential confounding factors, all participants were interviewed and completed a detailed questionnaire covering standard demographic characteristics, personal and family medical history, lifestyle habits, and prior IR exposure. For the purpose of smoking status, former smokers who had quit smoking at least 2 years prior to diagnosis were considered as non-smokers. Clinical and pathological examination was also performed.
Peripheral blood samples were collected from each patient into both 10 mL heparinized tubes (for cytogenetic analysis) and citrated tubes (for genotype analysis). For cytogenetic analysis, blood samples were drawn (1) prior to 131 I administration as well as 1, 6, and 24 months after therapy in patients submitted to a 70 mCi dose and (2) prior to 131 I administration as well as 1 and 3 months afterward in patients submitted to a 100 mCi dose. For genotype analysis, blood samples were stored at −80 • C until further use.
All subjects gave their informed consent for inclusion before they participated in the study. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of Instituto Português de Oncologia Francisco Gentil (GIC/357) and by the Ethics Committee of Faculdade Ciências Médicas (CE-5/2008).

Genotype Analysis
Genomic DNA was isolated from blood samples using the commercially available QIAamp ® DNA mini kit (QIAamp ® DNA mini kit; Qiagen GmbH, Hilden, Germany), according to the manufacturer's recommendations. The fluorimetric Quant-iT™ Picogreen ® dsDNA Assay Kit (Invitrogen, Waltham, MA, USA) was used to quantify and ensure uniformity in DNA concentration (2.5 ng/µL). DNA samples were kept at −20 • C until further use. SNPs were selected from those already analyzed by our team in a cohort of 106 DTC patients, according to selection criteria published elsewhere [18][19][20][21]. Due to sample size limitations, only SNPs presenting a minor allele frequency (MAF) > 0.15 in the original pool of patients were considered. MLH3 rs175080 was excluded a posteriori for insufficient genotype frequency (n ≤ 1) in at least one of the 131 I dose groups (Table S1). Overall, a total of 9 DNA repair SNPs across 3 DNA repair pathways (HR, NHEJ, and MMR) were considered for further analysis (Table 1). Table 1. Selected SNPs and detailed information on the corresponding base and amino acid changes, minor allele frequency, and Applied Biosystems (AB) assay used for genotyping. Genotyping was performed mostly by real-time polymerase chain reaction (RT-PCR): amplification and allelic discrimination were carried out on a 96-well ABI 7300 Real-Time PCR system thermal cycler (Applied Biosystems; Thermo Fisher Scientific, Inc., Waltham, MA, USA), following the manufacturer's instructions, with the use of the commercially available TaqMan ® SNP Genotyping Assays (Applied Biosystems) identified in Table 1. For XRCC3 rs861539 (HR pathway), genotyping was performed by conventional PCR-restriction fragment length polymorphism (RFLP) techniques. Primer sequences, PCR, and digestion conditions as well as expected electrophoretic patterns have been described [19]. To confirm genotyping and ensure accurate results, inconclusive samples were reanalyzed and genotyping was repeated in 10-15% of randomly chosen samples, with 100% concordance.

Cytogenetic Analysis
The cytokinesis-block micronucleus assay (CBMN) was used to analyze DNA damage and conducted according to standard methods. The methodology was performed and published as described previously [22][23][24]. The frequency of binucleated cells carrying micronuclei (BNMN), defined as the number of cells with MN per 1000 binucleated lymphocytes, is expressed as a count per thousand (% ). The Cytokinesis-Block Proliferation Index (CBPI) was determined according to the formula CBPI = [MI + 2MII + 3(MIII + MIV)]/N, where MI-MIV correspond to the number of human lymphocytes with one to four nuclei, respectively, and N is the total number of cells analyzed.

Statistical Analysis
All analyses were done with SPSS 22.0 (IBM SPSS Statistics for Windows, version 22.0, IBM Corp, Armonk, NY, USA) except for deviation of genotype distributions from Hardy-Weinberg equilibrium (HWE) and linkage disequilibrium (LD) analysis between SNPs on the same chromosome, which were performed with SNPstats [25].
Categorical variables, presented as frequencies and percentages, were compared between dose groups and with the original cohort of DTC patients by the Pearson's Chi-square (χ 2 ) test or the two-sided Fisher's exact test whenever 2 × 2 contingency tables were possible. For continuous variables (BNMN frequency, CBPI, and their net variation from baseline), presented as mean ± standard deviation, the normality and homogeneity of variances were evaluated by the Shapiro-Wilk and Levene tests, respectively. Longitudinal comparisons were performed by the paired sample t test (whenever a normal distribution could not be excluded) or the Wilcoxon signed-rank test (remaining cases) while the parametric Student t test (normal distributions) or the nonparametric Mann-Whitney U test (non-normal distributions) for independent samples were used for cross-sectional comparisons between the two 131 I dose groups and between different gender, age class, smoking status, histological type of tumor, and genotype categories.
Variable transformation was considered, when practically useful: DTC patients were dichotomized according to age, with the cut-off point being defined as the median age of all patients included (54 years). Due to limited sample size (hence, low frequency of homozygous variant genotypes), a dominant model of inheritance was assumed for all SNPs. Moreover, the net variation in BNMN frequency (i.e., therapy-induced BNMN) was calculated by subtracting the background (pre-treatment) BNMN frequency from the corresponding post-treatment values. This is an exploratory 'proof of concept' study, not a conclusive final one. As such, the Bonferroni adjustment was deemed as not necessary as it is too conservative. Furthermore, the complement of the false-negative rate β to compute the power of a test (1-β) was not taken into account at this stage since larger studies are needed to change this preliminary study into a confirmatory one. Statistical significance was set at p < 0.05.

Characteristics of the Study Population
A general description of the study population is presented in Table 2. The age of DTC patients submitted to 131 I therapy ranged from 32 to 73 years, with a mean of 52.54 ± 11.62 years. As expected, female patients (88.5%, n = 23) greatly outnumbered male patients (11.5%, n = 3) and papillary carcinoma cases (PTC, 69.2%, n = 18) were also more frequent than follicular ones (FTC, 30.8%, n = 8), in agreement with gender and histotype distributions commonly reported for DTC [1,2,4]. Overall, 15.4% (n = 4) of patients were smokers. No significant differences in patient age, gender, histological type of tumor, and smoking status were observed between groups submitted to different 131 I doses ( Table 2) nor between any of these groups (separated or together) and our original DTC population [18].

Cytogenetic Data
The frequency of BNMN (mean ± S.D.) in the 26 DTC patients submitted to 131 I therapy and included in this study is illustrated in Figure 1 and summarized in Table S2. Pre-treatment and post-treatment values are presented, stratified by dose group.

Cytogenetic Data
The frequency of BNMN (mean ± S.D.) in the 26 DTC patients submitted to 131 I therapy and included in this study is illustrated in Figure 1 and summarized in Table S2. Pre-treatment and posttreatment values are presented, stratified by dose group. The results from the 70 mCi dose group until 6 months after 131 I administration have been published before [22]. As it was not possible to collect genotyping data on 4 of the original 19 patients, these patients were excluded and the data were re-analyzed. Longitudinal results in this dose group are, nevertheless, similar to those originally reported [22]: as evident from Figure 1, BNMN frequency in these patients increases significantly 1 month after 131 I therapy (from 5.27 ± 3.63‰ to 8.80 ± 4.65‰, p = 0.039) and stabilizes at 6 months after 131 I therapy (8.93 ± 5.92‰, p = 0.944 vs. 1 month after therapy), remaining persistently higher than before treatment (p = 0.041).
To investigate the long-term persistence of such therapy-induced damage, the study of these patients at 2 years after therapy was extended (Table S2 and Figure 1). Cytogenetic data at such time point was available for 11 patients only. The frequency of BNMN remained stable (9.64 ± 2.80‰, similar to values at 1 and 6 months, p = 0.460 and p = 0.328, respectively) and persistently higher than baseline (p = 0.005).
To confirm these findings and check for a possible dose effect, the study was replicated in an independent group of patients administered with 100 mCi. As expected, BNMN frequency was significantly higher in the 100 mCi group than in the 70 mCi group, irrespective of the time point (Table S2 and Figure 1), suggesting a dose-effect association (hence, a cause-effect relation) between iodine dose and BNMN levels. Apart from this quantitative difference, the effect of either dose on BNMN frequency was qualitatively similar, BNMN in the 100 mCi group increasing significantly 1 The results from the 70 mCi dose group until 6 months after 131 I administration have been published before [22]. As it was not possible to collect genotyping data on 4 of the original 19 patients, these patients were excluded and the data were re-analyzed. Longitudinal results in this dose group are, nevertheless, similar to those originally reported [22]: as evident from Figure 1, BNMN frequency in these patients increases significantly 1 month after 131 I therapy (from 5.27 ± 3.63% to 8.80 ± 4.65% , p = 0.039) and stabilizes at 6 months after 131 I therapy (8.93 ± 5.92% , p = 0.944 vs. 1 month after therapy), remaining persistently higher than before treatment (p = 0.041).
To investigate the long-term persistence of such therapy-induced damage, the study of these patients at 2 years after therapy was extended (Table S2 and Figure 1). Cytogenetic data at such time point was available for 11 patients only. The frequency of BNMN remained stable (9.64 ± 2.80% , similar to values at 1 and 6 months, p = 0.460 and p = 0.328, respectively) and persistently higher than baseline (p = 0.005).
To confirm these findings and check for a possible dose effect, the study was replicated in an independent group of patients administered with 100 mCi. As expected, BNMN frequency was significantly higher in the 100 mCi group than in the 70 mCi group, irrespective of the time point (Table  S2 and Figure 1), suggesting a dose-effect association (hence, a cause-effect relation) between iodine dose and BNMN levels. Apart from this quantitative difference, the effect of either dose on BNMN frequency was qualitatively similar, BNMN in the 100 mCi group increasing significantly 1 month after therapy (from 9.64 ± 4.78% to 17.27 ± 5.14% , p = 0.011) and remaining persistently higher than baseline at 3 months (21.40 ± 5.66% , p < 0.001 and p = 0.054 compared to pre-treatment and 1 month post-treatment values, respectively) (Table S2).
Moreover, of notice, the BNMN increment (net balance) after 131 I therapy was more pronounced in the 100 mCi group than in the 70 mCi group, despite the difference was not significant (p > 0.05).
Finally, the CBPI (mean ± S.D.) was also determined for the 15 DTC patients submitted to therapy with 70 mCi 131 I. As depicted in Figure 2, this index, which indicates the proliferation capacity of lymphocytes and may be used to calculate cytotoxicity [26], did not change appreciably at 1 and 6 months after 131 I administration but was markedly reduced at 24 months after therapy (from 1.78 ± 0.13 to 1.53 ± 0.09, p = 0.001).
Moreover, of notice, the BNMN increment (net balance) after 131 I therapy was more pronounced in the 100 mCi group than in the 70 mCi group, despite the difference was not significant (p > 0.05).
Finally, the CBPI (mean ± S.D.) was also determined for the 15 DTC patients submitted to therapy with 70 mCi 131 I. As depicted in Figure 2, this index, which indicates the proliferation capacity of lymphocytes and may be used to calculate cytotoxicity [26], did not change appreciably at 1 and 6 months after 131 I administration but was markedly reduced at 24 months after therapy (from 1.78 ± 0.13 to 1.53 ± 0.09, p = 0.001).

Characteristics of the Study Population and Cytogenetic Data
The potential influence of the demographic, lifestyle, and clinical characteristics of the study population on cytogenetic data was also evaluated. As depicted in Figure 3, in patients treated with 70 mCi, histology interfered with both pre-treatment BNMN levels and its net balance 1 month after 131 I therapy ( Figure 3): basal BNMN frequency was significantly higher in FTC than in PTC patients (8.20 ± 3.11‰ vs. 3.80 ± 3.01‰, p = 0.020) but, 1 month after therapy, increased only in PTC patients, resulting in a significantly different net balance between the two histotypes (+6.20 ± 5.05‰ in PTC vs. −1.80 ± 3.96‰ in FTC, p = 0.009). Such effect was not observed in 100 mCi-treated patients nor when both dose groups were considered together. Likewise, no significant effect of gender, age, or smoking habits on BNMN levels or its net balance was detected, irrespective of the time point or dose group. Furthermore, except maybe for gender, no significant effect on CBPI was observed for any of these variables in the 70 mCi dose group. Baseline CBPI values were borderline higher in female compared to male patients (p = 0.045) but such finding should not be overvalued as only one male patient was included in this dose group.

Characteristics of the Study Population and Cytogenetic Data
The potential influence of the demographic, lifestyle, and clinical characteristics of the study population on cytogenetic data was also evaluated. As depicted in Figure 3, in patients treated with 70 mCi, histology interfered with both pre-treatment BNMN levels and its net balance 1 month after 131 I therapy ( Figure 3): basal BNMN frequency was significantly higher in FTC than in PTC patients (8.20 ± 3.11% vs. 3.80 ± 3.01% , p = 0.020) but, 1 month after therapy, increased only in PTC patients, resulting in a significantly different net balance between the two histotypes (+6.20 ± 5.05% in PTC vs. −1.80 ± 3.96% in FTC, p = 0.009). Such effect was not observed in 100 mCi-treated patients nor when both dose groups were considered together. Likewise, no significant effect of gender, age, or smoking habits on BNMN levels or its net balance was detected, irrespective of the time point or dose group. Furthermore, except maybe for gender, no significant effect on CBPI was observed for any of these variables in the 70 mCi dose group. Baseline CBPI values were borderline higher in female compared to male patients (p = 0.045) but such finding should not be overvalued as only one male patient was included in this dose group.   Table 3 reports the allele frequency and genotype distribution of 9 DNA repair SNPs among our sample of 131 I-treated patients. Genotype distributions were consistent with HWE in either dose group or their combination (p > 0.05) and, except for MSH3 rs26279, did not differ significantly from those described in our previously studied DTC population ( c ). For MSH3 rs26279, non-uniform distribution was observed, with the common allele being overrepresented in the study sample compared to the original population (p = 0.048, in the dominant model, Table S1). Moreover, importantly, no significant differences in genotype distributions were detected between dose groups, for any of the SNPs, irrespective of the model of inheritance assumed (Table 3). No relevant linkage  Table 3 reports the allele frequency and genotype distribution of 9 DNA repair SNPs among our sample of 131 I-treated patients. Genotype distributions were consistent with HWE in either dose group or their combination (p > 0.05) and, except for MSH3 rs26279, did not differ significantly from those described in our previously studied DTC population ( c ). For MSH3 rs26279, non-uniform distribution was observed, with the common allele being overrepresented in the study sample compared to the original population (p = 0.048, in the dominant model, Table S1). Moreover, importantly, no significant differences in genotype distributions were detected between dose groups, for any of the SNPs, irrespective of the model of inheritance assumed (Table 3). No relevant linkage association was observed between any of the SNPs.

DNA Repair SNPs and Cytogenetic Data
The influence of DNA repair SNPs on BNMN frequencies and the corresponding variation from pre-treatment values is shown in Figure 4, Table 4, Table 5 and Tables S3-S5. Prior to 131 I administration, BNMN frequency was higher in patients carrying the MLH1 rs1799977 variant allele than in those homozygous for the common allele, with the difference being significant in the 100 mCi dose group (p = 0.012) and in the pool of both groups (p = 0.019). MAF, minor allele frequency. All comparisons of genotype distributions were performed by the twosided Fisher's exact test (whenever 2 × 2 contingency tables are possible) or the χ 2 test (remaining cases). No significant differences among the 70 and 100 mCi dose groups were observed.

DNA Repair SNPs and Cytogenetic Data
The influence of DNA repair SNPs on BNMN frequencies and the corresponding variation from pre-treatment values is shown in Figure 4, Table 4, Table 5 and Tables S3-S5. Prior to 131 I administration, BNMN frequency was higher in patients carrying the MLH1 rs1799977 variant allele than in those homozygous for the common allele, with the difference being significant in the 100 mCi dose group (p = 0.012) and in the pool of both groups (p = 0.019).     131 I dose group at t 0 , t 1 , t 3 /t 6, and t 24 , according to genotype (only SNPs presenting significant findings are shown).   One month after 131 I administration, MLH1 rs1799977 variant allele carriers always presented significantly lower BNMN levels than patients homozygous for the common allele, either when considering absolute values (p = 0.004, p = 0.012 and p = 0.034 in the 70 mCi, 100 mCi, and in the pool of both groups, respectively) or the net variation from baseline (p = 0.002, p = 0.001 and p < 0.001 in the 70 mCi, 100 mCi and in the pool of both groups, respectively). BNMN frequency one month after therapy was also significantly lower in carriers of the variant allele for NBN rs1805794 (p = 0.043 in the 100 mCi group and p = 0.017 in the pool of both groups), with the difference in net BNMN values almost being significant (p = 0.099 in the 100 mCi dose group and p = 0.058 in the pool of both groups). Further, carriers of at least one MSH4 rs5745325 variant allele exhibited higher levels of 131 I-induced BNMN than patients homozygous for the common allele (p = 0.018 in the 100 mCi group, p = 0.043 in the combination of both groups), with the difference in absolute BNMN frequencies being significant in the pooled analysis of both groups (p = 0.039) and almost significant in the 100 mCi group (p = 0.084).
Three months after therapy, significantly higher BNMN frequencies were found in patients from the 100 mCi group carrying the MSH3 rs26279 variant allele (p = 0.030).
No other significant difference in either absolute or therapy-induced BNMN frequencies was found between the different genotypes of the DNA repair SNPs, at any time point. Likewise, no influence of genotype in CBPI, either absolute or relative to baseline values, was detected for any of the DNA repair SNPs considered in this study, at any time point (Table S6).

Discussion
We have previously demonstrated a significant increase in BNMN frequency in peripheral lymphocytes from 19 DTC patients treated with 70 mCi 131 I [22]. In the present exploratory study, in order to confirm these findings, to evaluate the long-term persistence of such 131 I-induced DNA damage and to determine whether it may be influenced by DNA repair SNPs, we extended our analysis at 2 years after 131 I administration in this group of patients, included a new group of patients submitted to RAI therapy with 100 mCi and profiled 9 DNA repair SNPs in patients from both groups.
In line with our previously reported results, we observed, in the 100 mCi dose group, a significant and persistent increase in BNMN frequency after 131 I therapy, with mean levels being always higher than in the 70 mCi group, irrespective of the time point considered. Replication across two independent sets of patients and observation of a dose effect strongly suggests a causal relation between RAI therapy and systemic chromosomal damage in lymphocytes, as assessed by the MNCB assay. Such correlation has been repeatedly demonstrated (both in thyroid patients following RAI therapy [27][28][29][30][31][32] and in other settings where exposure to low levels of low-LET (linear energy transfer) ionizing radiation occurs [28,33]) and is expected since 131 I may be taken up by extra-thyroidal cells [7] and emit βand γ-radiation capable of inducing dose-dependent chromosomal damage detectable by cytogenetic analysis (e.g., micronuclei) [27,28,32]. The ability of 131 I to induce cytogenetic damage in peripheral lymphocytes in a dose-dependent manner is, in fact, clear and well-established, allowing BNMN frequency to be used as a valid, highly sensitive, and specific biomarker of effect for biological dosimetry of RAI therapy and, hence, to predict its associated genotoxic risk in dividing mammalian cells [27,28,32,34,35].
A less clear picture exists, however, concerning the long-term persistence (kinetics of the recovery) of such IR-induced cytogenetic damage. Our results from the 70 mCi dose group suggest that 131 I-induced damage in peripheral lymphocytes persists for at least 2 years. Despite negative results have also been published [36,37], our results are in line with most prior follow-up studies on RAI therapy or other low-dose IR exposures (e.g., for diagnostic purposes) [28,29,[38][39][40][41]. Considering the half-life of 131 I (ranging from 1 to 8 days in thyroidectomized and non-thyroidectomized TC patients, respectively) [28] and of circulating lymphocytes (about 3 years) [28,38], such repeated demonstration of persistent cytogenetic damage is somehow surprising and challenge the widely held views about the mechanisms of IR-induced DNA damage. Possible explanations for the long-term genomic instability of lymphocytes from 131 I-exposed subjects include the introduction, upon irradiation, of DNA damage and cytogenetic alterations (1) in a subset of long-lived naïve T lymphocytes, quiescent cells that survive for prolonged periods of time in a resting stage, retaining the initially inflicted DNA damage and expressing it as micronuclei when stimulated to proliferate in the CBMN assay [38,42,43], (2) in hematopoietic stem and progenitor cells that, through clonal expansion, may give rise to mature T lymphocytes with stable and unstable aberrations, perpetuating genomic instability in time (transgenerational effect) [38,42,43], and (3) in non-irradiated lymphocytes (a delayed non-targeted effect), as a result of the long-term production and plasma secretion of soluble clastogenic factors by irradiated cells (oxidative stress by-products such as ROS (reactive oxygen species) and inflammatory cytokines such as TNF-α) that may further extend IR-induced cytogenetic damage in time ("bystander effect") [44]. The two latter explanations are generally favored, as a large number of studies exist demonstrating either the high frequency of gene mutations and chromosomal aberrations in the progeny of irradiated cells or the production and plasma release of factors with clastogenic activity by irradiated cells (including one on 131 I-treated patients) [37]. Overall, current evidence [44][45][46][47] supports the notion that a potent long-term inflammatory-type response develops upon IR exposure, irradiated cells producing danger signals (oxidative stress by-products and inflammatory cytokines) capable of exerting an array of persistent bystander effects in non-irradiated cells (altered levels of damage-inducible and stress-related proteins), leading to delayed genomic instability (chromosomal aberrations, sister chromatid exchanges, micronuclei formation/induction or mutations), hence, predisposing to malignancy (altered proliferation or transformation). Such long-term inflammatory-type response could also be responsible for the marked reduction in CBPI that we observed at 24 months after 131 I therapy.
In this study, complying with current recommendations, we also investigated the role of potential confounding factors on BNMN frequency. As reviewed elsewhere [48][49][50] and demonstrated through meta-analysis in the International Human MicroNucleus (HUMN) Project [51], age and gender are well-established factors, with increasing age and female gender being consistently associated with higher BNMN levels in peripheral blood lymphocytes. The influence of age has been demonstrated, in particular, in 131 I-treated patients [28,31]. Data on the potential role of smoking status on BNMN levels are somewhat more inconsistent, and many studies failing to find an association except, maybe, in heavy smokers and in those with relevant occupational exposures [48][49][50][51]. In this study, no significant effect of gender, age, or smoking habits on BNMN levels or its net balance was detected, irrespective of the time point or dose group. The study was probably underpowered to detect such effects. It is also possible that the effect of these variables may have been masked by the impact of internal IR exposure after 131 I administration.
We did observe, however, in the 70 mCi group only, differences on BNMN levels between the two TC histotypes, as FTC patients presented significantly higher basal BNMN frequency than PTC patients but significantly lower therapy-induced BNMN levels at one month after 131 I administration. This is suggestive of higher background genomic instability in FTC but higher sensitivity to the DNA damaging effects of IR in PTC. Considering the small sample size and the non-reproducibility of the findings between the two dose groups, extreme caution must be taken in the interpretation of these results. Nevertheless, the available evidence supports both findings: PTC usually presents as a microsatellite stable tumor, with no appreciable levels of either loss of heterozygosity (LOH) or aneuploidy (stable chromosome profile) [52][53][54]. On the contrary, a considerable degree of chromosomal instability appears to be a hallmark feature of FTC, which presents a consistently higher frequency of chromosomal abnormalities, LOH, allelic loss, and a higher mutational burden compared to PTC [52,53,[55][56][57]. Microsatellite instability (MSI), despite uncommon in TC, also appears to be more frequent in FTC than in PTC [53][54][55]. The available evidence thus largely supports our observation of higher background genomic instability in FTC. Moreover, considering that activating RAS mutations are commonly observed in FTC but not in PTC [53,58,59], the association between increased RAS expression and decreased frequency of IR-induced MN reported by Miller et al. [60] is coherent with our own observation of lower 131 I-induced BNMN frequency in FTC, supporting the idea that this histotype is less sensitive to the DNA damaging effects of IR than PTC. Such hypothesis (i.e., higher sensitivity to IR in PTC) is further reinforced by a recent observation, through meta-analysis, of increased efficacy of RAI therapy in PTC patients, compared to FTC [61] but more studies are needed for a solid conclusion to be drawn.
Moreover, in the present study, we further evaluated the potential impact of selected HR, NHEJ, and MMR pathway SNPs on BNMN levels, before and after the administration of 131 I. To our knowledge, this is the first study doing so. Significant genotype effects on MN frequency and/or its net balance were observed for HR (NBN) and MMR (MLH1, MSH3, MSH4) repair pathway SNPs across different time points. This was expected because (1) IR exposure results in increased DNA damage, most notably, single-and double-strand breaks, oxidative lesions (e.g., 8-oxoG), DNA-protein crosslinks (DPCs) and clustered DNA lesions [62][63][64][65][66][67]; (2) the HR pathway, acting in the S/G2 stages of the cell cycle, is the major DNA repair pathway involved in the error-free correction of DSBs [11,33,35,68]; (3) MMR proteins, besides their canonical actions on the post-replication repair of mispaired nucleotides and insertion-deletion loops, have also been demonstrated to play an important role on the damage response to IR-induced DSBs, either through cooperation with HR or through signaling for cell-cycle arrest and apoptosis [64,[69][70][71]; (4) DSBs, if left unrepaired, e.g., due to the presence of SNPs that reduce the DNA repair capacity, may give rise to chromosome breakage and MN formation upon replication [28,33,35,72]. The potential influence of functional DSB repair SNPs on 131 I-induced BNMN frequency is, therefore, fully justified. A literature review on the functional impact of these SNPs and their putative association with response to radio and/or chemotherapy was performed and is presented below ( Table 6). Table 6. Literature review on the functional impact of the studied SNPs and their putative association with radio and/or chemosensitivity (only SNPs presenting significant findings in the present study are shown).
GG genotype associated with increased radiosensitivity in cancer patients, translating into increased efficacy [78] or toxicity [79] of radiotherapy (alone or combined with chemotherapy).

MSH3 rs26279
Missense SNP located in the ATPase domain, critical for protein activity [80]; altered expression has been suggested [81] but not confirmed [82].
GG genotype associated with decreased incidence of radiation dermatitis in breast cancer patients receiving radiotherapy [83], decreased overall survival in head and neck squamous cell carcinoma patients submitted to radiochemotherapy [81] and decreased response to platinum-based chemotherapy in advanced non-small cell lung cancer patients [84].

MSH4 rs5745325
Missense SNP located in the N-terminal domain, involved in the interaction with eIF3f [85].
None to be reported.
MLH1, together with PMS2, forms the MutLα heterodimer, a complex critical for the maintenance of genomic integrity [103,104]. The common rs1799977 (c.665A>G, Ile219Val) missense SNP is located in a region that codes for a highly conserved N-terminal ATPase domain, vital for MLH1 function. However, since both alleles code for nonpolar pH-neutral amino acids, the substitution is considered conservative and not expected to result in drastic changes in protein properties and function [73].
Several functional studies support this hypothesis [73,74,[105][106][107] but the existence of a more subtle effect should not be excluded [73,106,108,109] as an association between the G variant allele and reduced MLH1 expression has been demonstrated repeatedly in cancer patients [74][75][76][77]. Moreover, two recent meta-analyses have associated this variant with increased risk of colorectal cancer [110,111]. Considering the important role that MLH1 plays in the maintenance of genome integrity and cancer avoidance, both observations are compatible with our own observation of increased baseline BNMN levels in TC patients carrying the G allele. A different picture emerges, however, upon IR exposure: as previously stated, MMR proteins such as MLH1 play a dual role in the DNA damage response to IR, triggering cell-cycle arrest and allowing for either DSB repair or apoptosis [11,64]. MMR proficiency is thus expected to result in higher repair efficiency of IR-induced damage (hence, lower cytogenetic levels) and, simultaneously, higher cytotoxicity upon IR exposure (hence, increased sensitivity to radiotherapy). Indeed, alongside with increased cancer susceptibility, the MLH1 rs1799977 variant GG genotype has been associated with increased radiosensitivity in cancer patients, translating into increased efficacy [78] or toxicity [79] of radiotherapy (alone or combined with chemotherapy). This is suggestive of increased MMR proficiency in such patients and supports our own observation of significantly lower BNMN levels, one month after 131 I therapy, in TC patients carrying the G allele. How the same allele may be associated with decreased function under basal conditions and increased function after IR exposure remains to be explained: MLH1 has been demonstrated to be upregulated upon IR exposure [112,113], it is possible that such upregulation might be more pronounced in G allele carriers, but this is highly speculative. Nevertheless, the high level of significance in our observations (especially when considering the change in MN frequency from baseline) and their cross-validation in independent groups strengthen our conclusions and warrant further studies to clarify this issue.
Two other MMR polymorphisms presented significant findings in our study, MSH3 rs26279 and MSH4 rs5745325. Like MLH1, MSH3 also appears to be involved in the repair and damage response to IR-associated lesions such as DSBs and inter-strand crosslinks [84,114]. MSH3 rs26279 (c.3133A>G; Thr1045Ala) is a common SNP that results in an amino acid change in the ATPase domain of MLH3. This domain is critical for MSH3 activity, suggesting a functional impact for this variant [80]. Such hypothesis remains to be verified as, to the best of our knowledge, functional studies are lacking. An association with altered MSH3 expression levels has been suggested [81] but not confirmed [82]. The MSH3 rs26279 G allele or GG genotype has been consistently associated with cancer risk in all 3 meta-analysis that we are aware of, particularly for colon and breast cancer [115][116][117], suggesting decreased DNA repair capacity in G allele carriers. Further, MSH3 rs26279 GG homozygosity has also been associated with decreased incidence of radiation dermatitis in breast cancer patients receiving radiotherapy [83], decreased overall survival in head and neck squamous cell carcinoma patients submitted to radiochemotherapy [81], and decreased response to platinum-based chemotherapy in advanced non-small cell lung cancer patients [84], suggesting decreased sensitivity to DNA damaging agents such as IR or platinum in GG homozygous individuals. Such phenotype is commonly associated with MMR deficiency [64,69,70,118,119]. If we consider, once again, the dual role that MMR proteins such as MSH3 play in damage repair and apoptosis, these results are compatible with decreased G allele function, resulting in decreased DNA repair and apoptosis, increased damage tolerance, resistance to radio/chemotherapy, and reduced efficacy and cytotoxicity of such therapeutic agents. Our own observation of increased MN levels in TC patients carrying the G allele, 6 months after receiving 100 mCi 131 I, fits comfortably into this picture.
Likewise, in our study, MN frequency was also significantly increased (absolute and change from baseline values) in TC patients carrying the A allele of MSH4 rs5745325, one month after 131 I administration. MSH4 rs5745325 (c.289G>A; Ala97Thr) has only seldom been evaluated: on single SNP analysis, two prior studies by our team failed to detect an association with either thyroid [21] or breast cancer risk [120]. The same was observed in the only two other association studies that we found focusing on this SNP [121,122]. Interestingly, in three out of these four studies, significant associations were detected when interactions with other SNPs-MSH6 rs1042821 [21], MLH3 rs175080 [120], and CHRNA5 rs16969968 [121]-were considered. Besides the important role that MSH4 plays in recombinational repair during meiosis [123], it is also suggested to participate, through interaction with a vast array of binding partners, in DSB-triggered damage response and repair [85,123,124]. It is possible that MSH4 rs5745325 interferes with the binding properties of MSH4, with impact on its putative contribution to the DNA damage response and repair. The interaction of MSH4 with eIF3f (a subunit of the eIF3 complex implicated in apoptosis regulation and tumor development), for example, occurs at the region comprising the first 150 amino acids of the N-terminal domain of MSH4 (where rs5745325 is located) and has been demonstrated to foster hMSH4 stabilization and to modulate sensitivity to IR-induced DNA damage [85]. This is in line with our own findings.
Finally, we also observed a significant association between NBN rs1805794 and BNMN frequency, one month after the administration of 100 mCi 131 I. Nibrin plays a pivotal role in the initial steps of the cellular response to DNA damage, directly initiating DSB repair through the RAD51-dependent HR pathway and further contributing to cell cycle checkpoint activation through an ATM-dependent pathway [68,[125][126][127]. Inactivating germline mutations in the NBN gene (which encodes for the Nibrin protein) markedly impair DSB repair and cause the Nijmegen breakage syndrome, characterized by chromosomal instability, increased cancer susceptibility, and increased sensitivity to DSB-causing agents such as IR or cisplatin. These features highlight the importance of Nibrin for genome stability (hence, cancer prevention) [86,93,125,127]. NBN overexpression also appears to be associated with poor prognosis in several types of cancer [68], which is consistent with a putative increase in DNA repair efficiency, hence, resistance to cytotoxic therapy. Among the numerous NBN polymorphisms, rs1805794 (c.553G>C; Glu185Gln) is the most frequently investigated. This missense variant results in an amino acid change in the BRCT (BRCA1 C Terminus) domain (amino acids , a domain involved in the interaction of Nibrin with BRCA1. The resulting complex (the BRCA1-associated genome surveillance complex, BASC) is responsible for the recognition and repair of aberrant DNA [86][87][88][89]. NBN rs1805794 has been suggested to interfere with the interaction properties of Nibrin and thus with DNA repair capacity, sensitivity to DNA damaging agents (such as IR) and cancer susceptibility. Accordingly, NBN rs1805794 has been repeatedly associated with cancer risk, as demonstrated by numerous meta-analysis [68,88,89,125,[128][129][130][131][132] but conflicting reports exist [126,127,133,134]. Interestingly, the association may vary according to ethnicity [88,130] and tumor site [125], as one of these meta-analysis has demonstrated, for example, increased risk of leukemia, nasopharyngeal, and urinary system cancers but decreased risk of lung, gastric, and digestive system cancers [125]. Furthermore, final conclusive evidence on the significance of NBN rs1805794 is still lacking, as the functional studies performed thus far have yielded negative or conflicting results: while lymphocytes from healthy individuals homozygous for the G allele have been reported to present higher DNA damage levels (as assessed by the Comet assay) than lymphocytes from C allele carriers [90], opposite results have been reported in ex vivo X-ray irradiated cells from healthy subjects [88]. Further ex vivo irradiation studies have failed to observe a significant influence of NBN rs1805794 on DNA repair capacity and radiosensitivity [91,92]. Furthermore, since a putative functional impact of this SNP on DNA repair capacity could possibly influence patient sensitivity to radio and/or chemotherapy, association studies correlating NBN rs1805794 genotype with therapy response, toxicity, or prognosis have also been performed. Again, most studies failed to find an association in radiotherapy [79,[93][94][95][96] or chemotherapy [97][98][99] treated patients, while other studies presented opposite findings, associating the NBN rs1805794 C allele with either improved [86,100] or worse [68,101] prognosis upon platinum-based chemotherapy. Interestingly, increased frequency of binucleated lymphocytes with nucleoplasmic bridges was observed in peripheral lymphocytes from children with high environmental exposure to IR that were heterozygous for NBN rs1805794, while the reverse patter was observed in children homozygous for the Gln allele [102]. This may be suggestive of molecular heterosis, a hypothesis that, considering the high interethnic variability of the NBN rs1805794 distribution, could help in explaining such divergent results. Overall, despite extensively investigated, the functional significance of NBN rs1805794, as well as its putative role in sensitivity to DNA damaging agents (such as IR) and cancer susceptibility remains elusive, warranting further studies to clarify this issue.

Conclusions
In conclusion, our results confirm that BNMN levels in peripheral lymphocytes from DTC patients increase significantly immediately 1 month after 131 I therapy and further suggest that these remain stable and persistently higher than baseline for at least 2 years. Furthermore, a marked reduction in CBPI is observed at 24 months after 131 I administration. Moreover, HR and MMR SNPs (MLH1 rs1799977, MSH3 rs26279, MSH4 rs5745325, and NBN rs1805794) were, for the first time, associated with IR-induced MN, a cytogenetic marker of DNA damage, in TC patients submitted to 131 I therapy. Among such findings, a highly significant and independently replicated association was observed for MLH1 rs1799977, strongly suggesting a role for this particular SNP on the personalization of RAI therapy in TC cancer patients. Baseline and post-therapy MN levels also diverged according to tumor histotype. These results should be regarded as merely suggestive and proof of concept, as the sample was small and the number of tests was high, increasing the likelihood of false-positive results. Nevertheless, our findings suggest that TC therapy with 131 I may pose a long-term challenge to cells other than thyrocytes and that the patient genetic profile may influence the individual sensitivity to this therapy. Such hypotheses are of relevance to the efficacy and safety of 131 I therapy, a widespread practice in TC patients. As such, extending the benefit already achieved with the latest guidelines on TC treatment in terms of risk/benefit ratio through improved clinical assessment of the potential long-term risks of 131 I therapy is desirable. Likewise, despite the micronucleus test is considered the gold standard methodology in genetic toxicology testing and often used as a "stand-alone" test in numerous and relevant papers in this area, other tests should also be employed to validate these results. Furthermore, potential radiogenomic markers such as those suggested here should be evaluated in larger samples, preferentially through multi-center independent studies adequately powered to provide more robust evidence and, eventually, to allow for gene-gene and gene-environment interactions to be assessed. Identifying the most clinically relevant variables, genetic or non-genetic, and accurately estimating their impact on 131 I therapy response rate and adverse event risk for each individual TC patient is the ultimate goal, under a personalized medicine approach.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4425/11/9/1083/s1, Table S1: Allele and genotype frequencies in thyroid cancer patients submitted to 131 I therapy (n = 26) and in the original (reference) DTC population (n = 106), Table S2: BNMN frequency (% , mean ± S.D.) in DTC patients before and after (1, 3/6, and 24 months) therapy with different doses of 131 I (70 and 100 mCi), Table S3: Frequency of micronucleated cells (% BNMN, mean ± SD) in the 70 mCi dose group at t 0 , t 1 , t 6 and t 24 , and corresponding variation, according to genotype, Table S4: Frequency of micronucleated cells (% BNMN, mean ± SD) in the 100 mCi dose group at t 0 , t 1 and t 3 , and corresponding variation, according to genotype, Table S5: Frequency of micronucleated cells (% BNMN, mean ± SD) in the combined dose groups at t 0 and t 1 , and corresponding variation, according to genotype, Table S6: Cytokinesis-Block Proliferation Index (CBPI, mean ± SD) in the 70 mCi dose group at t 0 , t 1 , t 6 and t 24 , and corresponding variation, according to genotype.  Acknowledgments: The authors warmly acknowledge the generous collaboration of patients and controls in this study as well as of our colleague Ana Paula Azevedo for technical support.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.