Background
In nuclear medicine, treatment using radiolabeled tumor-seeking agents is a vital option for patients with, e.g., metastases or inoperable tumors [
1].
211At is a synthetic α-emitter with metalloid and halogen characteristics and a mean linear energy transfer (LET) value that is nearly optimal for inducing double-strand breaks (DSB) in DNA [
2,
3]. As such,
211At has a high biological effectiveness for cell killing which renders it a promising candidate for radiolabeling in radionuclide therapy [
3-
7]. However,
211At can be liberated from a carrier molecule
in vivo during degradation or metabolism and be retained in normal tissues [
8-
10]. Unbound
211At is accumulated in the thyroid gland through a mechanism similar to iodide resulting in high uptake and absorbed dose [
9-
15]. In other tissues,
211At uptake occurs at lower levels, although generally higher than for iodide, with tissue-specific differences leading to differential exposure to ionizing radiation throughout the body [
8-
10]. The increased uptake of
211At in non-thyroid tissues compared to, e.g., iodide increases the radiation risk and side effects in healthy tissues after treatment with
211At-labeled tumor-targeting agents. A detailed understanding of normal tissue responses is needed to establish accurate tolerance doses and thus to optimize treatment effectiveness. Basic research in nuclear medicine routinely uses nude mice to study tumor xenografts and, for instance, anti-tumor effects of targeted therapy. However, little is known about the quality and quantity of normal tissue effects
in vivo, which is an important aspect for robust analysis and establishment of both effective and safe treatment parameters.
Many studies for biomarker discovery of ionizing radiation effects have been performed with
in vitro model systems and external irradiation, predominantly γ-rays and X-rays [
16,
17]. In contrast, knowledge on basic tissue responses
in vivo to internal radionuclide exposure is still scarce, specifically with regard to α-emitters and low absorbed dose. Within the framework of the European project Low Dose Research towards Multidisciplinary Integration (DoReMi), Pernot and colleagues summarized that ‘changes in RNA levels identified by transcriptomics’ had ‘unknown sensitivity’ and that ‘specificity to ionizing radiation and confounders (was) unknown at present time’ [
18]. In previous studies on mice, we used intravenously administered
211At in the 0.064- to 42-kBq range and characterized genome-wide transcriptional responses after 24 h in various normal tissues: in thyroid [
19] and in kidney cortex and medulla, liver, lungs, and spleen [
20]. Radiation-induced cellular responses were demonstrated as complex and tissue-specific and to vary with absorbed dose level in a non-linear manner, which advises caution for extrapolation or interpolation of responses between absorbed dose levels [
19,
20]. Furthermore, comparatively few previously proposed ionizing radiation-associated genes from
in vitro studies were differentially regulated, which demonstrated the need to identify biomarker genes in an
in vivo setting. Despite pronounced differences in absorbed dose levels between thyroid and non-thyroid tissues, the total significant transcript regulation showed similar characteristics between all tissues, i.e., a distinct shift in regulation intensity between 0.64 and 1.8 kBq
211At. These findings suggested that responses in non-thyroid tissues were not only due to low-dose effects from
211At but also subject to systemic effects from, e.g., the
211At-accumulating thyroid gland [
20]. These studies highlight the necessity to further expand the knowledge base on low-dose responses to ionizing radiation
in vivo while regarding induced effects in a systemic and physiological context.
In this exploratory study, transcriptional gene expression responses were analyzed on a genome-wide scale using RNA microarray technology. Microarrays are a sophisticated method for both hypothesis building and hypothesis testing within the same experimental setup due to the vast amount and large scale of obtained data without limitation to a set of presupposed genes or regulatory pathways. However, to the best of our knowledge, only a few studies have been performed investigating basic transcriptional responses to internal radionuclide exposure to, e.g.,
131I [
21,
22] or specifically
211At [
19,
20]. Furthermore, few robust molecular biomarkers for ionizing radiation exposure
in vivo have been identified, since most studies have been performed
in vitro [
16,
17].
The purpose of this study was to investigate genome-wide transcriptional responses over time in normal non-thyroid tissues following intravenous 211At administration in mice. The aim was to characterize tissue-specific transcriptome responses and to identify potential in vivo biomarkers for (very) low mean absorbed doses with sensitivity to dose rate.
Discussion
The synthetic radiohalogen
211At and the
211Po daughter emit two main α-particles, i.e., 5.87 and 7.45 MeV, which have an approximated range of 48 and 70 μm in liquid water, respectively [
25]. Alpha particles are densely ionizing and cause non-homogeneous energy deposition within cells or tissues. The degree of homogeneity in cellular irradiation - i.e., the fraction of non-hit, single-hit, or multi-hit cells in a population - most likely influences the biological response of a given tissue. This response depends not only on the amount of administered activity and tissue-specific uptake and clearing rates but also on, e.g., organ morphology, tissue structure, and size and density of cells ([
29], Josefsson A, Forssell-Aronsson E. Microdosimetric analysis of the radio halogens
123I,
124I,
125I,
131I and
211At, Submitted). Non-homogeneous irradiation may result in non-linear response to absorbed dose on the tissue level and/or increased intensity of, e.g., non-targeted effects [
30]. Microdosimetric considerations of
211At exposure
in vivo and the relevance of non-homogeneous irradiation at very low to low absorbed doses have been discussed previously in a related study [
20]. Accordingly, very low to low absorbed doses in kidney, liver, lungs, and spleen tissues corresponded to non-homogeneous irradiations with large fractions of non-hit cells. With increased absorbed dose, the fraction of single-hit and multi-hit cells increases. In this context, it should be pointed out that genome-wide transcriptional responses were obtained from homogenized tissue samples. The presented data gives information on significant biological responses within tissue-specific cell populations in an
in vivo context. Nevertheless, the non-homogeneous nature of α-particle exposure at (very) low absorbed doses from
211At introduces the question how much of the significant transcript response originated from the (single or multi) hit fraction(s) and to what extent the non-hit fraction contributed to the observed effects in each tissue.
The total number of significantly regulated transcripts in response to 1.7 kBq
211At varied with time and tissue type. Previous studies demonstrated that differential transcript expression in thyroid and non-thyroid tissues varied with
211At activity at 24 h after administration [
19,
20]. Taken together, these findings substantiate for the
in vivo setting that ionizing radiation-induced responses strongly depend on exposure condition and time course. Recently, Heinonen and colleagues estimated differential expression time periods after stress exposure using a novel Bayesian likelihood ratio test [
31]. Among others, the group measured transcript profiles of 77 differentially expressed genes in primary human endothelial cells
in vitro after external irradiation of 2 Gy with a
137Cs source, demonstrating that the maximum differential expression (>1.5-fold change) occurred between 8 and 12 days after irradiation. Time-dependent change in differential expression constitutes an important factor in transcriptional analysis and, specifically, biomarker discovery. In the present study using internal radionuclide exposure, differences in differential expression between time points were influenced not only by, e.g., the biological response time of certain regulatory events but also by continuous irradiation and increased absorbed dose over respective periods. Hence, it is unfeasible to analyze time between exposure and effect as an individual factor in this setting. An external irradiation setup would be required to allow for both singular and continuous radiation exposure in order to differentiate between biological response time and (extent of) induced regulation.
In the dose-rate study, injected activities of 105 and 7.5 kBq were chosen so that the thyroid gland would receive a mean absorbed dose of 1.4 Gy over 1 and 6 h, respectively. Since the thyroid is a regulatory organ that is presumed to have a potential impact on non-thyroid tissue responses [
20], we chose this setup to keep absorbed dose in thyroid tissue constant over both time periods. This would constrict the variable of absorbed dose dependence in the systemic setting of a regulatory organ impacting target organs. Concerning non-thyroid tissues, tissue-specific variations in
211At uptake and clearing rates did not give exact matching of mean absorbed dose within this setup, but respective mean absorbed doses lay at similar low levels. Nevertheless, transcript regulation showed sensitivity to differential dose rate exposure in this setup for all tissues. These regulatory responses also showed pronounced differences compared with thyroid tissue, which is subject of an ongoing study.
An inverse dose rate effect on total transcript regulation was observed in the kidney cortex, liver, and lungs. The inverse dose rate phenomenon has been demonstrated for different exposure conditions and biological end points, for instance for mutation induction in human lymphoblasts in response to
137Cs γ-rays [
32], deletions size in the
HPRT locus in human lymphoblastoid cells following 200 keV X-ray exposure [
33], micronuclei induction in Lewis lung carcinoma cells exposed to
60Co γ-rays [
34], and cytokine gene expression in human glioblastoma cell lines irradiated with
137Cs γ-rays [
35]. A cell cycle-dependent phase of sensitivity has been proposed to explain the inverse dose rate effect, meaning that increased irradiation time at decreased dose rate increases the probability of a cell to enter a sensitive phase during cell-cycle progression [
36,
37]. A saturable intermediate state has been hypothesized as well, which postulates that an increase in probability for damage induction would not increase proportionally with increased number of hits to a cell [
37,
38]. Brenner and colleagues concluded in their modeling approach using the linear-quadratic + resensitization formalism that ‘all potential explanations of inverse dose rate effects predict that, at appropriately low doses, no dose rate effects of any kind are expected’ [
38]. This statement provokes the question on the nature of the inverse dose rate effects observed for differential transcript expression at the (very) low absorbed doses from α-radiation observed in this study. Assuming relatively homogeneous tissue morphology and radionuclide distribution, one out of one thousand cells is estimated to be hit in the mGy range with very low probability for more than one hit per cell. Since the largest amount of cells contributing to expression data are within the non-hit fraction under these exposure conditions, complex non-targeted effects may contribute to non-linearity of responses. Furthermore, the relatively high LET value and biological effectiveness of
211At-emitted α-particles may increase the inverse dose rate effectiveness factor compared with γ-rays and X-rays. It is unclear at this point whether the transcripts that exhibited an inverse dose rate effect are regulated as an immediate response to particle hits or if their regulation is determined by regulatory networks in dependence of additional factors. The finding that dose rate effects differed between tissues implies an additional degree of complexity for radiation responses to (very) low absorbed doses or dose rates
in vivo. Biological response on the organismic level needs to be understood in a dynamic context where some tissues would respond more intensively at a given time point or to a certain dose rate, while other tissues would show a less pronounced response; yet this relation might change or invert over time and with regard to dose rate.
A biomarker for ionizing radiation exposure should be indicative of a certain radiation-induced effect, such as molecular damage to the DNA or impaired cellular function. An ideal biomarker should be easily quantifiable and constitute a molecule (or a group of molecules) that changes in concentration or is modified upon exposure. The response should change monotonously with absorbed dose and occur early upon irradiation and remain detectable after a long time period. Sensitivity over a wide absorbed dose range is desirable but not mandatory, since a panel of biomarkers could be assorted covering a wider range. A biomarker should be sensitive to low dose rate and, ideally, indicate low, medium, or high dose rate exposure with, e.g., changes in time-of-onset and intensity of response. Transcriptional regulation responds to a stressor in a sensitive manner and continuous dysregulation of transcript expression can be expected to manifest itself earlier than detrimental changes on the protein level. This is an advantage of transcriptional biomarkers compared to protein biomarkers or general markers of tissue function. In a low dose exposure setting, an acute radiation response and related tissue damage are not expected, as opposed to an increased risk for cancerogenesis or latent damage. Correlation between early effects upon exposure and long-term health effects requires substantial knowledge of overall transcriptomic responses and individual gene regulation patterns. This approach can be used to establish risk estimation based on absorbed dose or, in critical cases, justify tissue biopsies of risk organs.
In this study, we investigated the up- or downregulation of transcript expression for the discovery of individual molecular biomarkers. BALB/c nude mice were chosen for these experiments since
in vivo models for radionuclide therapy require immunodeficient mice to establish human tumor xenografts. The presented results should be applicable in that analytical context when studying normal tissue effects of radiolabeled agents in tumor-bearing mice. In this context, it should be noted that reduced expression of the DNA-dependent protein kinase catalytic subunit (DNA-PKcs) and single nucleotide polymorphisms (SNPs) in
Prkdc have been reported for several BALB/c strains which affected DNA repair proficiency upon ionizing radiation exposure [
39-
43]. To the best of our knowledge, the specific strain used in this study has not been categorized regarding DNA repair deficiency, but it is assumed that the genetic background exhibits similarly increased radiation sensitivity. Accordingly, the observed responses can be related to radiation sensitive individuals in a clinical setting.
For the kidney cortex, liver, lungs, and spleen, several genes were identified that followed a potentially indicative pattern across all exposures, i.e., a consistent regulatory change between lower and higher injected activity at 1 and 6 h with either a direct or inverse dose rate effect - while showing a consistent trend relative to 1.7 kBq over time. Several of these genes were shared between tissues: angiopoietin-like protein 4 (
Angptl4) was regulated in both the kidney cortex and liver and was nearly regulated in all tissues and at all mean absorbed doses (0.064 to 42 kBq
211At) after 24 h in the previous study [
20]. The
Angptl4 gene product is thought to modulate vascular activity and tumor cell motility and invasiveness, which implies its significance for tumor progression and metastasis [
44-
46]. The prognostic value of
Angptl4 deregulation in ionizing radiation-induced cancerogenesis is promising, but accumulating evidence suggests that the function of
Angptl4 highly depends on, e.g., proteolytic processing and posttranslational modifications and thus can have opposite effects on vascular permeability in different cancers [
47,
48]. The versatile functions of
Angptl4 impede further speculation on long-term effects after
211At exposure, or ionizing radiation in general, at this point. Long-term studies are needed in order to evaluate whether or not the observed response is indicative of radiation-induced tumorigenesis or carcinogenesis at these early time points or if transcriptional regulation of
Angptl4 responds in a different context.
Either
Per1 or
Per2 (periodic clock genes 1 and 2) were commonly regulated in the kidney cortex, liver, and lungs at all exposure conditions. In our previous study on non-thyroid tissues,
Per1 also showed significant upregulation in these tissues after 24 h in an absorbed dose range from several mGy to around 1 Gy, i.e., specifically after administration of 0.64, 14, and 42 kBq
211At (data not shown in Langen et al. [
20]; please refer to GEO:GSE40806). In thyroid,
Per1 was also consistently upregulated after 24 h at these injected activities, i.e., 0.5, 11, and 32 Gy, respectively (see supplemental material of Rudqvist et al. [
19]).
Per2, on the other hand, was significantly regulated only in liver and thyroid after 24 h in the low injected activity range. In both tissues,
Per2 was consistently downregulated with a similarly low differential expression ranging between −1.7 and −2.2-fold change. Taken together, we demonstrated that
Per1 and
Per2 responded differently in different tissues and across a wide-absorbed dose regimen from mGy to over 32 Gy.
Per genes encode for negative regulators in the circadian feedback loop, thus regulating metabolism as well as various other cellular processes and circadian-dependent gene expression [
49]. Disturbance of the circadian clock has been linked to cancer; specifically, deregulation of
Per1 and
Per2 have been connected to gastric cancer and suggested as prognostic markers [
49-
51]. Moreover, long-term effects in tumor suppression by Per proteins have been demonstrated in context with ionizing radiation exposure [
52]. At a sub-lethal dose of 4 Gy γ-radiation,
mPer2 mutant mice showed graying of the coat after approximately 2 months and/or developed lymphoma after 5 months at a significantly increased rate compared with wild-type mice [
52]. Consequently, disruption of
Per gene expression can be expected to dampen - if not abolish - impede this protective feature and result in increased cancer risk. Whether responses in
Per1 and
Per2 are indicative of radiation-induced malignancies or if their regulation is mainly affected by metabolism and circadian rhythm needs to be addressed in further studies allowing for strict separation of these factors.
Tsc22d3 (TSC22 domain family, member 3) was upregulated in the liver, lungs, and spleen at all treatment conditions. This finding was in agreement with a study by Koike and colleagues demonstrating upregulation of
TSC22 in normal human epidermal keratinocytes after 4 and 8 h following exposure to 10 Gy X-ray radiation [
53]. In our previous study,
Tsc22d3 did not exhibit significant regulation after 24 h in liver or kidney tissues, and only few regulation incidences in lungs and spleen (data not shown in Langen et al. [
20]; please refer to GEO:GSE40806).
Tsc22d3 was also not significantly regulated after 24 h in thyroid across an absorbed dose range from 0.05 to 32 Gy (see supplemental material of Rudqvist et al. [
19]). In comparison with the aforementioned
Per transcripts,
Tsc22d3 appeared to respond in a more restrictive fashion. The expressed protein shares sequence similarities with leucine zipper proteins, implying a function as a transcription factor [
54]. Regulation of
Tsc22d3 appears to have a key function in anti-inflammatory and immunosuppressive effects of glucocorticoids and interleukin-10 [
55]. Furthermore, interactions between
Tsc22D3 and transcription factors nuclear factor NF-kappa-B p105 subunit (NFKB1) and nuclear factor NF-kappa-B p100 subunit (NFKB2) have been demonstrated [
55]. The knowledge base, however, is still scarce regarding long-term effects of
Tsc22d3 deregulation, and prognosis of late effects from
211At-induced upregulation cannot be made at this point. In a broader sense, even robust molecular biomarkers would underlie regulation networks, which would render extrapolation of effects over time difficult. Hence, observed responses in individual genes should be considered in the broader context of regulatory networks within a cell and in the larger context of tissues and organs within the body.
Cellular function (the quality of transcriptional effects) was characterized according to enriched transcript-associated biological processes. The response patterns differed between tissues at similar absorbed dose level - specifically comparing kidney cortex, kidney medulla, and liver - which was in agreement with a study demonstrating that ionizing radiation-induced effects are not preprogrammed genetic responses but rather depend on tissue origin [
56]. The major lesion type of ionizing radiation is considered to be the induction of DSB in the DNA molecule. The LET value of
211At-emitted α-particles is nearly 100 keV/μm with a high effectiveness of producing DSB [
2,
3], which should be considered in the analysis and interpretation of induced effects. In the present study, DNA damage and repair pathways did not show transcriptional response in enriched biological processes. A fold change value of at least 1.5 was chosen to exclude a large proportion of weakly responding genes and identify more pronounced changes in regulation, and accordingly, more suitable candidate genes (transcripts). In another research context, a somewhat lower fold change threshold might be used which would show more weakly responding genes (transcripts). Another factor that may lead to false-negative observation is data convolution from mixed cell populations, i.e., significant responses in a certain cell type may be dampened due to low cell type frequency. Nevertheless, when the DNA damage burden does not exceed the level of DNA damage recognition, transcriptional regulation of respective proteins is not expected. Accordingly, responses in DNA damage and repair processes may not be sensitive biomarkers in the low-dose regimen. However, chromatin organization was strongly affected after 1 h in kidney medulla and also showed response to at least one absorbed dose level after 24 h in all of the investigated tissues in the previous study [
20]. Based on studies by Bakkenist and Kastan, chromatin organization was speculated to be a potential biomarker for ionizing radiation exposure [
57]. The work indicated that activation of the ataxia-telangiectasia mutated (ATM) protein - a crucial player in DNA damage recognition - potentially results from structural changes of chromatin organization and may not strictly depend on direct protein-DNA binding [
57]. In fact, responses in chromatin organization were detected in each of the investigated tissues after 24 h at certain (very) low absorbed doses from
211At as previously reported [
20]. These findings suggest monitoring of biological processes for chromatin organization, i.e., respective key genes involved in process regulation, as potential
in vivo biomarkers for (very) low absorbed doses of ionizing radiation. In this regard, chromatin organization may exhibit increased biomarker sensitivity for α-emitters compared to β-emitters due to the higher probability for producing DSB lesions per decay event.