Gene expression signature in mouse thyroid tissue after 131I and 211At exposure

Background 131I and 211At are used in nuclear medicine and accumulate in the thyroid gland and may impact normal thyroid function. The aim of this study was to determine transcriptional profile variations, assess the impact on cellular activity, and identify genes with biomarker properties in thyroid tissue after 131I and 211At administration in mice. Methods To further investigate thyroid tissue transcriptional responses to 131I and 211At administration, we generated a new transcriptional dataset that includes re-evaluated raw intensity values from our previous 131I and 211At studies. Differential transcriptional profiles were identified by comparing treated and mock-treated samples using Nexus Expression 3.0 software. Further data analysis was performed using R/Bioconductor and IPA. Results A total of 1144 genes were regulated. Hierarchical clustering subdivided the groups into two clusters containing the lowest and highest absorbed dose levels, respectively, and revealed similar transcriptional regulation patterns for many kallikrein-related genes. Twenty-seven of the 1144 genes were recurrently regulated after 131I and 211At exposure and divided into six clusters. Several signalling pathways were affected, including calcium, integrin-linked kinase, and thyroid cancer signalling, and the peroxisomal proliferator-activated receptor network. Conclusions Substantial changes in transcriptional regulation were shown in 131I and 211At-treated samples, and 27 genes were identified as potential biomarkers for 131I and 211At exposure. Clustering revealed distinct differences between transcriptional profiles of both similar and different exposures, demonstrating the necessity for better understanding of radiation-induced effects on cellular activity. Additionally, ionizing radiation-induced changes in kallikrein gene expression and identified canonical pathways should be further assessed.


Background
Medical applications for radionuclides are rapidly developing. The β particle-emitting 131 I is frequently included in therapy regimens of various thyroid disorders due to selective uptake of the isotope in thyroid tissue and is also administered bound to tumour-seeking agents for therapeutic and diagnostic purposes [1][2][3][4]. The α particleemitting 211 At is a suitable therapeutic radionuclide due to, e.g., a nearly optimal therapeutic linear energy transfer value of emitted α particles of 98.8 keV/μm [5]. 211 At-la-belled tumour-seeking pharmaceuticals have been utilized both in humans and in animals [6][7][8]. Similar to unbound 131 I, selective uptake of unbound 211 At also occurs in thyroid tissue [9][10][11], and administration of free 131 I and 211 At or 131 I-and 211 At-labelled radiopharmaceuticals have been shown to result in thyroid irradiation [6,12]. Additionally, nuclear accidents often involve an atmospheric release of 131 I, as was the case in connection with the Chernobyl accident, which resulted in an increased incidence of thyroid cancer in children [13,14].
Despite the risk of exposing thyroid tissue to 131 I and 211 At, the understanding of radiation-induced effects is far from complete and molecular biomarkers of absorbed dose or radiation-induced effects on thyroid tissue are yet to be identified. Biomarkers are useful to indicate achieved therapeutic effects or estimate risk exposure and evaluate the quality and severity of side effects. RNA microarray analysis is a semi-quantitative method to identify changes in genome-wide transcriptional patterns between two or more samples. The result is a transcriptional profile, i.e. a snapshot of the radiation-induced cellular activity at the mRNA level. This can be used to determine the impact of radiation on biological functions and canonical pathways, to predict upstream regulation of target molecules, and for biomarker discovery without the risk of bias in focusing on a specific set of signalling pathways only.
Few investigations on global gene expression effects of α and β particle irradiation have been performed in normal (thyroid) tissue in vivo. There are, however, a few in vitro studies on global gene expression in fibroblasts and cancer cells after α particle exposure [15,16]. Additionally, effects on gene expression for a set of pre-defined genes after α particle irradiation have been measured both in cancer cells in vitro and in vivo in xenografted tumours [17][18][19]. Previously, we have published results showing substantial differences between transcriptional profiles in thyroid tissue in vivo after different 131 I or 211 At exposures, varying absorbed dose, dose rate and time after administration [20][21][22]. We then identified potential biomarkers for each type of exposure separately and concluded that biological response to radiation is complex and that it is difficult to predict or extrapolate radiation-induced effects for other exposure parameters.
The aim of this work was to re-evaluate the previously obtained transcriptional response in thyroid after administration of 131 I − and free 211 At in mice from different exposure conditions to gain a better understanding of variations in transcriptional regulation on absorbed dose, dose rate, time after administration and radiation quality.

Study design
In this study, we further investigate the thyroid transcriptional response to 131 I and 211 At exposure by using normalized intensity values from three different experiments where separate analyses of each experiment have been published elsewhere [20][21][22]. All normalized intensity values (normalization was performed according to experiment) from these three experiments were imported together into Nexus Expression 3.0 (BioDiscovery; El Segundo, CA) for filtering, linear modelling and determination of differentially expressed genes.
No new tissue sampling and analysis was thus performed in this study, but a reanalysis of all these data together in a new analysis by Nexus Expression. Data presented in the present work are therefore novel and have not been previously published.
A brief summary of the methods used in the three original studies is as follows: a total of 44 female Balb/c nude mice (CAnN.Cg-Foxn1nu/Crl, Charles River Laboratories International, Inc., Salzfeld, Germany) (n = 2-3 per group) were i.v. injected with various amounts of 131 I or 211 At in the tail vein and killed at 1, 6, 24 or 168 h after administration, or mock-treated (Table 1). Thyroid, kidney, liver, lung and spleen tissue samples were collected and immediately snap-frozen in liquid nitrogen and stored at −80°C until RNA extraction. The present study contains data on transcriptional changes in thyroid glands. The transcriptional response in the kidneys, livers, lungs and spleens has been published elsewhere [23][24][25].
The methods for determination of absorbed doses have been reported previously with separate analysis of each experiment [20][21][22]. In short, the mean absorbed dose was calculated according to conventional Medical Internal Radiation Dose (MIRD) formalism. We used previously published data on relative activity concentration of 131 I and 211 At in thyroid in mice [26], an absorbed fraction of 0.742 and 1 for radiation emitted from 131 I and 211 At, respectively [27], and a standard mouse thyroid mass of 3 mg.

Gene expression analysis
Genome-wide transcriptional analysis using RNA microarray of thyroid tissue has been described elsewhere [20,21]. Briefly, RNA samples were analysed using MouseRef-8 Whole-Genome Expression Beadchips (Illumina; San Diego, CA, USA). Nexus Expression 3.0 (BioDiscovery; El Segundo, CA) was used to identify statistically significant differentially expressed transcripts (≥1.5-fold change) with a Benjamini-Hochberg adjusted p value cut-off of 0.01 between irradiated and control tissues. RNA microarray data from irradiated samples were compared with RNA microarray data from control animals from each separate experiment to ensure that differential expression of genes reflects radiation-induced changes and not variations between different control groups. There was one difference in study design between the present and previous studies. In the present work, data from tissue samples from animals administered 0.064 and 0.64 kBq 211 At and killed 24 h after administration were compared with data from tissue samples from control animals killed the same day. However, in the previous paper with analysis of the transcriptional response at 24 h after 211 At administration, animals administered 0.064 and 0.64 kBq were compared with controls killed earlier (killed simultaneously as animals injected with 1.7, 14 and 42 kBq, also at 24 h after administration but on another day) [20]. In the present paper, the term "regulated transcripts/genes" is used synonymous to "statistically significant differentially expressed transcripts/genes". Hierarchical clustering of regulated transcripts according to their transcriptional regulation profile was performed using the hclust function (stats package, version 3.1.1) with the complete linkage algorithm and Lance-Williams dissimilarity update formula in the R statistical computing environment (version 0.97.551, http://www.r-project.org) [28]. Heat maps were produced using the heatmap.2 function (gplots package, version 2.14.2).
Upstream regulation, diseases and functions and canonical pathway analyses were generated using the Ingenuity Pathway Analysis tool (IPA, Ingenuity® Systems, www.ingenuity.com; Redwood City, CA) with Fisher's exact test (p value <0.05).

Regulated genes
In the present study, 1144 genes (1164 transcripts) were regulated. The number of regulated genes in each group varied between 55 (17 Gy, 24 h, 131 I) and 630 (1.4 Gy, 1 h 211 At) ( Table 1). Hierarchical clustering subdivided the exposure groups into two larger clusters: one smaller branch containing groups with higher absorbed dose levels (1.4-32 Gy from 211 At and 8.5 Gy from 131 I at 24 h) and a larger branch with the remaining groups ( Fig. 1).
At both 1 h and 6 h after 211 At administration, a higher absorbed dose/dose rate resulted in a higher number of regulated genes (Table 1). In contrast, at 24 h after 211 At administration, the number of regulated genes varied non-monotonously with absorbed dose. A slight increase in the number of genes regulated after 131 I exposure was seen between 0.85 and 8.5 Gy and a decrease between 8.5 and 17 Gy. Transcriptional profiles at 1, 6 and 168 h clustered together, with the exception of 1.4 Gy at 1 h ( Fig. 1). At 24 h, the transcriptional profiles for 0.05 and 0.5 Gy clustered together while the profiles for 1.4, 11 and 32 Gy clustered together with highest similarity between 11 and 32 Gy. Furthermore, upregulation of 110 genes was shared between 0.05 and 1.4 Gy at 24 and 1 h after 211 At administration, respectively. For 131 I exposure, the transcriptional profiles for 0.85 and 17 Gy clustered together, whereas the response after 8.5 Gy was more similar to that after 1.4, 11 and 32 Gy at 24 h following 211 At administration. In the groups exposed to 1.4 Gy, the number of regulated genes decreased from 630 to 290 between 1 and 6 h, but increased to 359 at 24 h (Table 1). Additionally, the transcriptional profiles of these groups showed little similarity in the cluster analysis ( Fig. 1). Lastly, an equal amount of injected activity (1.7 kBq, 211 At) resulted in 210, 170, 359 and 136 regulated genes at 0.023, 0.32, 1.4 and 1.8 Gy after 1, 6, 24 and 168 h, respectively ( Table 1). The transcriptional profiles of these groups were similar and clustered together with the exception of 1.4 Gy at 24 h which was clearly different (Fig. 1).

Recurrently regulated genes after 131 I and 211 At exposure
Twenty-seven of the 1144 regulated genes were regulated in ≥9 of the 13 groups (Fig. 3). Hierarchical clustering divided the 27 genes into six groups: 1) Atp2a1, Ckm, Eno3, Pvalb, Tnnc2, Tnni2, Tnnt3; 2) Coq10b, Ctgf, Dbp, Per1; 3) Mfsd2; 4) Ltf; 5) Ccl8, Ly6g6d, S100a8, S100a9 and 6) Aoc3, Ccl9, Clec2d, Cpa3, Dpt, Fstl1, Lum, Mup2, Ogn, Scara3. In cluster 1, genes were downregulated at 1 h following 211 At administration. This changed at 6 h where 0.32 and 1.4 Gy showed downand upregulation, respectively. At 24 h, the regulated genes were upregulated and, notably, the fold change increased with absorbed dose. At 24 h after 131 I administration, all of these genes were upregulated with high fold changes (21-180) after 8.5 Gy, while two genes were upregulated to low extent after 17 Gy. Genes in cluster 2 were upregulated after 211 At administration while 131 I administration resulted in both up-and downregulation. The Mfsd2 gene, the only gene in cluster 3, was up-and downregulated after 211 At and 131 I administration, respectively. The Ltf gene, sole gene in cluster  Canonical pathway analysis: calcium, integrin-linked kinase and thyroid cancer signalling In the present study, calcium, integrin-linked kinase and thyroid cancer signalling were the top three canonical pathways generated using the IPA comparison analysis tool ( Table 2). An impact on calcium signalling was statistically significant in all groups except at the two lowest and the highest absorbed dose levels 24 h after 211 At and 131 I administrations, respectively. Integrin-linked kinase signalling was statistically significant in all groups except at low absorbed dose levels at 1 and 24 h, and at the lowest and highest absorbed dose levels 24 h after 211 At and 131 I administration, respectively. Genes associated with calcium and integrin-linked kinase signalling were generally downregulated early and at low absorbed doses and upregulated at later time points and at higher absorbed doses. Thyroid cancer signalling was statistically significant in all groups except at 0.023 and 0.05 Gy, 1 and 24 h following 211 At administration, respectively, and at 0.85 and 8.5 Gy 24 h after 131 I administration (Table 2). Additionally, the number of genes involved in the different canonical pathways at each time point generally increased with absorbed dose (with exceptions between 11 and 32 Gy).

Diseases and function analysis: thyroid cancer and disturbed thyroid function
A diseases and function analysis related to thyroid was generated with IPA (Table 3). At 6 (1.4 Gy) and 24 h (11 and 32 Gy) after 211 At administration, a relation between the transcriptional response and various thyroid cancer types was identified. Such a relation was also identified 24 h after 131 I administration; however, at a lower absorbed dose level (0.85 and 8.5 Gy). Additionally, the transcriptional response was linked to altered T3 and T4 levels and to thyroid gland development at 0.32 and 32 Gy, 6 and 24 h after 211 At administration, respectively.

Upstream regulation of molecules related to peroxisomal proliferator-activated receptors
The upstream regulation analysis generated by IPA predicted upstream regulation of various molecules related to peroxisomal proliferator-activated receptors (PPARs) ( Table 4). These included PPARA, PPARD, PPARG, PPARGC1A and several PPAR-targeting drugs and/or PPAR ligands such as GW501516, mono-(2-ethylhexyl)phthalate, pirinixic acid, Rosiglitazone and Troglitazone. IPA predicted that these upstream regulators were generally activated at 1 h, for absorbed dose ≤1.4 Gy at 24 h, and for 1.8 Gy at 168 h.

Discussion
In the present study, the values used to calculate absorbed dose were based on previously published 131 I and 211 At biodistribution data, where mice were simultaneously injected with both 131 I and 211 At, which allows for direct comparison of the absorbed dose per injected activity between 131 I and 211 At in the same animal [26]. In the present study, radioactivity measurements of individual thyroid samples would have enhanced the certainty in absorbed dose calculations but was not possible since all excised thyroid tissue was needed to ensure sufficient amount of RNA for microarray analysis. There are several important differences in the characteristics of 211 At and 131 I exposure: i) the difference in mean range of the α and β particles emitted (65 and 400 μm), ii) the much higher mean energy released per decay from 211 At compared with 131 I (7000 and 190 keV), iii) the difference in LET of particles emitted from 131 I and 211 At (0.25 and 98.8 keV/μm, respectively) and iv) much shorter half-life for 211 At than 131 I (7.2 h and 8.0 day, respectively). Taken together, 211 At irradiates more heterogeneously and with higher dose rates at similar absorbed dose levels compared with 131 I (and the dose rate will decline faster for 211 At compared to 131 I). The effects of radiation quality on global gene expression should be further studied, and to our knowledge, the present study is the first to investigate such differences between 131 I and 211 At.
RNA microarray analysis was used to evaluate the impact of 131 I and 211 At exposure on global transcriptional regulation in normal mouse thyroid tissue. Regulated genes were associated with biological functions using previously published literature reports and various databases, in addition to upstream and downstream regulation analysis and canonical pathway analysis generated by Ingenuity Pathway Analysis (IPA) software. In total, 1144 genes were differentially regulated showing a large variation in number of genes per group. In general, hierarchical clustering divided groups that received high absorbed doses into one branch and groups receiving low absorbed doses into another. Thus, we hypothesize that the transcriptional profiles presented here may reflect intrinsic biological properties predictive of 131 I and 211 At absorbed dose levels at various time points.
At both 1 and 6 h, the number of regulated genes increased with absorbed dose. This was not the case for 131 I or 211 At exposure at 24 h, where a broader range of   At administration were distinctly different although the absorbed dose was similar. This could in part be explained by the profound differences in dose rate; dose rate effects on the transcriptional response have previously been described in vivo following radionuclide administration [22,25,29]. These findings indicate that variations in the radiation-induced response with absorbed dose will be reflected in the number of regulated genes, in addition to which specific genes are regulated, although not in a clear dose-dependent manner. Instead it is likely that changes in transcriptional patterns in a specific tissue will depend with varying degree on, but not excluded to, the following parameters: exposure time, injected activity, absorbed dose, dose rate, dose distribution (e.g. frequency of non-, single-or multi-hit cells) and radiation quality. Each unique setup of these parameters may then yield a specific response in the target tissue. In addition, cells are dynamic systems with complex regulatory networks that activate cascades of downstream regulation that is sensitive to type and frequency of incoming stimulus.
It is valuable to identify genes with exposure-specific expression as they may be used as biomarkers. Biomarkers are useful to better understand the mechanisms behind the radiation-induced response. A potential application of biomarkers for ionizing radiation exposure of the thyroid might be in biological dosimetry after exposure to relatively high doses, maybe in a triage setting. In the present study, kallikrein 1 (Klk1) and 12 of 13 kallikrein 1-related (Klk1b) peptidases in the mouse genome were frequently regulated with fold change values between −3.8 and 110. The expression of these genes generally increased with absorbed dose and time after injection of 131 I or 211 At; however, 32 Gy resulted in less upregulation compared with 11 Gy (24 h, 211 At) and the highest dose rate used (1.4 Gy, 1 h, 211 At) resulted in downregulation. It is likely that the expression of Klk1 and Klk1-related peptidases depends, to a different degree, on dose rate, absorbed dose and time after injection. In a study on the rat urine proteome 24 h after 10 Gy total body irradiation, the occurrence of kallikrein 1-related peptidase b24 precursor protein increased while the kallikrein-binding serine protease inhibitor A3K precursor decreased [30]. In another study, the plasma kallikrein levels decreased with absorbed dose (0-19 Gy) at 2-24 h after local irradiation of the hind legs in tumour-bearing rats and controls [31]. Additionally, we have shown that regulation of Klk1 and Klk1-related genes in mouse thyroids after 131 I exposure does not show a circadian variation [32]. We hypothesize that genes involved in the kallikrein network may be potential biomarkers of radiation exposure, but further research is warranted to elucidate the relationship between radiation exposure and kallikrein proteases and kallikrein inhibitor levels. The kallikrein genes have also been shown to contribute to the radiation-induced death of various species. After treatment with soy bean trypsin inhibitors (SBTI), the mortality rate in mice and chickens 14 days after exposure to 690 and 820 R (6.7 and 8 Gy to soft tissue), respectively, decreased from 100 to 50 % in mice and from 86 to 4 % in chickens [33]. The authors suggested that the decrease in mortality rate after administration of Table 4 Peroxisomal proliferator-activated receptor (PPAR)-related upstream regulators in thyroids exposed to 131  A z score equal to or larger or less than 2 or −2 indicate activated or inhibited upstream regulator, respectively. A z score value between -2 and 2 is considered not statistically significant. n.s. indicates that IPA was not able to predict upstream regulation of that specific upstream regulator for the specific exposure condition SBTI originated from a radioprotective effect on the vascular system with less vascular leakage and that the protease inhibited was likely tissue pre-kallikrein. We suggest that the radioprotective role of SBTI should be further assessed.
Recurrently regulated genes might be potential biomarkers and show how different exposure types influence similar/related genes and biological functions, although maybe with different magnitude and/or direction of regulation. The 27 recurrently regulated genes in the present study were divided into six clusters according to the transcriptional pattern of each individual gene following a specific exposure. In cluster 1, 211 At-induced regulation was dependent on both absorbed dose and time after exposure with monotonous change in regulation at 24 h, and 8.5 Gy 131 I exposure resulted in very high upregulation. Genes in cluster 1 are related to muscular activity and/or calcium activity (Atp2a1, Eno3, Pvalb, Tnnc2, Tnni2 and Tnnt3). Notably, the thyroid gland contains parafollicular cells (C-cells) that produce the calcium homeostasis regulating hormone calcitonin. Cluster 2 contains genes related to various biological functions, e.g. cellular and tissue development and wound healing (Ctgf) [34], energy transduction (Coq10b) and circadian rhythm (Dbp, Per1) [35,36]. These genes were generally upregulated and may be indicators of radiation exposure in general. Cluster 3 consisted of only one gene (Mfsd2) that was up-and downregulated after 211 At and 131 I exposure, respectively, indicating a difference between radiation qualities (Ctgf, Per1, S100a8, S100a9, also showed a radiation quality dependency). No clear connection to the immune system, inflammation or the cytokine system was found for genes in clusters 1-3. However, the sole gene in cluster 4 (Ltf), all genes in cluster 5 (Ccl8, Ly6g6d, S100a8, S100a9) and a majority of genes in cluster 6 (Aoc3, Ccl9, Clec2d, Cpa3, Fstl1, Scara3) were related to the immune system in various ways, and many related to both inflammation and the cytokine network [37][38][39][40][41][42][43][44][45][46][47]. The cytokine encoded by Ccl9 is associated with systemic inflammation and has increased expression in macrophages after exposure to triiodothyronine [43]. Lactoferrin-in mice encoded by the Ltf gene, solely expressed in cluster 4 and generally downregulated-has been patented as a radioprotective drug and increased survival in mice exposed to 10 Gy (whole-body, external irradiation) via an impact on, e.g., cytokine regulation [38]. In cluster 5, genes were oppositely regulated when comparing 131 I and 211 At exposure, indicating a radiation quality-dependent immune response. In cluster 6, genes were downregulated at low absorbed doses and upregulated at high absorbed dose levels even though the shift from downto upregulation occurred at a lower absorbed dose level for 211 At compared with 131 I. This suggests that the radiation-induced regulation of genes in cluster 6 is dependent on both radiation quality and absorbed dose. The 27 recurrently regulated genes can potentially be used to discriminate between several different exposure parameters, e.g. radiation quality, absorbed dose levels and time after administration, and might be considered as potential biomarkers for at least 131 I and 211 At exposure of thyroid. These results indicate a connection between specific exposures and biological responses, especially for the genes in clusters 4-6 that were clearly associated with immunological response, inflammation and the cytokine network. These recurrently regulated genes should be further studied to better understand their impact on radiation-induced biological responses and in particular the local and systemic effects that involve inflammation, the immune system and the cytokine network.
To assess systemic effects from 131 I and 211 At exposure, regulation of the 27 recurring genes was compared with transcriptional changes in the lungs, spleen, liver and kidney cortex and medulla in the same mice dissected in the present study [23,24]. These non-thyroidal tissues, that are exposed at a much lower absorbed dose level compared with thyroid, shared regulation of 19/27 and 6/27 recurring genes after 211 At and 131 I exposure, respectively. Additionally, we have previously shown that the transcriptional response in the lungs, spleen, liver and kidney cortex and medulla in mice administered 131 I and 211 At can partly be explained as a systemic response from radiation-induced effects on thyroid [32]. One gene with potential biomarker properties is Dbp. The Dbp gene expression pattern changed in several non-thyroidal tissues after low absorbed dose level exposure to both 131 I and 211 At [23,24], in kidneys in mice both early and late after 177 Lu-octreotate administration [48], and in rat thyroids after 131 I administration [49].
A connection to thyroid cancer was detected for 5/27 recurring genes. PVALB has been suggested as an ideal biomarker to discriminate between benign and malignant thyroid cancer [50]. ENO3 is another cancer-related gene, associated to the PAX8-PPARG fusion protein in thyroid follicular carcinomas, and upstream regulation of PPARs and PPAR-related pathways was detected in the present study [51]. The level of CTGF correlated with metastasis, tumour size and clinical stage for papillary thyroid carcinoma in a previous study [52]. Furthermore, undifferentiated thyroid carcinomas have been shown to be S100A8/9 immunopositive and both genes were associated with, e.g., inflammation-associated cancer and aggressive breast cancer [41,53,54].
According to the Ingenuity canonical pathway and diseases and functions analysis tool, the transcriptional response after both 131 I and 211 At exposure was related to thyroid cancer signalling and various thyroid cancers, respectively. It is uncertain whether induction of thyroid cancer can be detected at the transcriptional level at these early times after initiation of radiation exposure. However, according to IPA, exposure to ionizing radiation activates thyroid cancer signalling by rearrangements of RET and/ or NTRK, both present in some of the exposed groups. Unfortunately, all parts of the thyroid samples from mice in the three studies this work is based on were used for microarray analysis, why further studies of genomic rearrangements were not possible from the same samples.
Additionally, KLK3 (human denotation of KLK1 in mouse) is among the involved molecules in all groups that show an impact on thyroid cancer signalling. Since IPA uses human protein nomenclature for annotation of genes, the presence of KLK3 in the IPA analysis is likely a result of regulation of mouse Klk1 and Klk1-related genes in the present study.
The number of annotated genes involved in calcium signalling generally increased with absorbed dose for 211 At exposure. Several of the genes associated with these molecules could be found in cluster 1 among the 27 recurrently regulated genes (Tnni2, Tnnc2, Tnnt3 and Atp2a1). Additionally, gene products of several other recurrently regulated genes are also calciumrelated according to literature reports, but not associated with calcium signalling in the IPA canonical pathway analysis. For example, the gene products of Clec2d and Ogn both inhibit osteoclasts that can release Ca 2+ into the blood [55,56]. As previously mentioned, the thyroid gland also contains parafollicular cells that produce calcitonin, a hormone partly responsible for calcium homeostasis and an inhibitor of osteoclast activity. However, the impact on calcitonin levels from 131 I and 211 At exposure was not investigated in the present study. A relationship between calcium and radiation-induced response has been previously reported and it was shown that calcium was required for bystander-induced apoptosis in unirradiated keratinocytes [57].
An impact on integrin-linked kinase (ILK) signalling was identified using an IPA canonical pathway analysis, and some genes associated with calcium signalling were also associated with ILK signalling in the present study. For 211 At, a higher absorbed dose generally resulted in a response involving a higher number of annotated genes. For 131 I exposure, ILK signalling was only statistically significant after 8.5 Gy, suggesting a difference in response due to radiation quality. In blood from mice administered with 137 Cs, genes associated with integrin-signalling were found upregulated at days 2 and 3 and downregulated at days 20 and 30 (transcriptional level) [58]. No clear temporal effect on integrin-signalling was seen during the somewhat shorter time range used in the present study. Interestingly, however, in the present study, we demonstrate that genes, e.g., associated with ILK signalling were generally upregulated at higher absorbed dose levels and downregulated at lower absorbed dose levels, suggesting different involvement of this signalling pathway at different absorbed dose levels. This was especially the case 24 h after 211 At administration, but a similar trend was also seen after 131 I administration. ILK signalling and radiation damage have been previously connected, and furthermore, ILK signalling partly controls cell adhesion and mediates prosurvival and antiapoptotic signalling after exposure to ionizing radiation [59]. Additionally, several cell adhesion GO terms were identified when performing in-depth separate analysis of the transcriptional response of thyroid tissue from animals in the three experiments that constitute the present study [20][21][22].
In the present study, the predicted upstream regulation of several peroxisomal proliferator-activated receptors (PPARs), and PPAR ligands and agonists was found. The PPARs are of interest in the radiationinduced biological response. In one study, administration of the PPARα agonist fenofibrate prevented some cognitive function impairment in young rats exposed to 40 Gy fractionated whole-brain irradiation [60]. In another study, knockout of PPARα resulted in inhibition of radiation-induced apoptosis in the mouse kidney through regulation of Nfkb and anti-apoptosis factors [61]. Together, these results indicate that the PPAR network may play a role in radiation-induced biological response and that it may be targeted to modulate radiation damage in various tissues.

Conclusions
A profound effect on gene expression in mouse thyroid tissue after 131 I and 211 At exposure was detected, and 27 genes, of which many are associated with immune response, were identified as potential biomarkers for 131 I and 211 At exposure, and the biomarker applicability of these genes should be further studied. Hierarchical clustering revealed distinct differences between transcriptional profiles of both similar and different exposures, demonstrating the necessity for better understanding of radiation-induced changes in cellular activity. Additionally, the kallikrein network deserves further attention since literature shows how administration of protease inhibitors, known to decrease kallikrein levels, drastically increased survival in various irradiated species. An effect on thyroid cancer signalling was identified, as well as regulation of several genes previously identified as biomarkers for thyroid cancer; however, it is unlikely that thyroid cancer can be manifested at the transcriptional level at these early times after initiation of radiation exposure. Furthermore, the present study supports that ionizing radiation may impact calcium-related biological processes. Taken together, we consider RNA microarrays, especially in the in vivo setting, to be an important tool to gain further insight in the complex radiationinduced changes in cellular activity and affected pathways, and for biomarker discovery.

Compliance with ethical standards
All applicable international, national, and/or institutional guidelines for the care and use of animals were followed. The study design was approved by the Ethical Committee on Animal Experiments in Gothenburg, Sweden. This article does not contain any studies with human participants performed by any of the authors.