Radiation-induced translational control of gene expression

ABSTRACT

Radiation-induced gene expression has long been hypothesized to protect against cell death. Defining this process would provide not only insight into the mechanisms mediating cell survival after radiation exposure, but also a novel source of targets for radiosensitization. However, whereas the radiation-induced gene expression profiles using total cellular mRNA have been generated for cell lines as well as normal tissues, with few exception, the changes in mRNA do not correlate with changes in the corresponding protein. The traditional approach to profiling gene expression, i.e., using total cellular RNA, does not take into account posttranscriptional regulation. In this review, we describe the use of gene expression profiling of polysome-bound RNA to establish that radiation modifies gene expression via translational control. Because changes in polysome-bound mRNA correlate with changes in protein, analysis of the translational profiles provides a unique data set for investigating the mechanisms mediating cellular radioresponse.

KEYWORDS:

• gene expression
• polysomes
• radiation
• translational control

Radiotherapy remains a major cancer treatment modality, with about two-thirds of patients receiving radiation sometime during their therapy.^1,2 Improving the efficacy of radiotherapy is thus likely to have a significant impact on the outcome of cancer treatment in general. Toward this end, one approach is to target the molecules participating in cellular radioresponse, a strategy that depends on the delineation of the fundamental mechanisms involved. The best defined mechanisms mediating eukaryotic cellular radiosensitivity focus on the post-translational modification of constitutively expressed proteins, such as those involved in the repair of DNA damage and the activation of cell cycle checkpoints. However, in prokaryotes it has long been recognized that radiation also modifies gene expression, which is considered a critical component of the adaptive response protecting against cell death.³ If a similar process occurs in eukaryotes, it would provide not only insight into the mechanisms mediating cell survival after radiation exposure, but also a novel source of targets for radiosensitization.

Along these lines, the radiation-induced changes of total cellular mRNA levels (i.e., the transcriptome) have been extensively cataloged using single gene and whole genome approaches. Radiation-induced gene expression profiles have been generated for a variety of cell lines as well as normal tissues as a function of dose and time after irradiation.^4-7 Although genes associated with potentially relevant pathways and processes have been identified, interpreting the significance of radiation-induced gene expression has been limited by 2 major factors. First, the changes in gene expression detected after irradiation are primarily cell line-dependent, even between lines isolated from the same tissue.^8-11 This suggests that radiation-induced gene expression is highly dependent on the individual cell genotype, as opposed to the tissue of origin or tumor type.¹² The second and probably the most important barrier to understanding the significance of radiation-induced gene expression is that changes in mRNA levels, with few exceptions,^13,14 do not reflect changes in the corresponding protein. This disconnect between mRNA and protein levels in irradiated cells was directly addressed by Szkanderova et al. who found no correlation between 10 proteins induced in L929 cells by radiation and their corresponding mRNAs.¹⁵ Consistent with an unclear significance of the radiation-induced transcriptome, work in yeast showed that genes transcriptionally upregulated by radiation had little to no effect on cell survival.¹⁶

The discrepancy between radiation-induced changes in mRNAs and their corresponding proteins is similar to that observed after other forms of stress, a situation attributed to the translational control of gene expression.^17,18 In 2003, Rajasekhar et al. reported that Ras and Akt, which are implicated in the regulation of cellular radiosensitivity, influenced the recruitment of existing mRNAs to polysomes, suggesting that these 2 proteins participate in translational control.¹⁹ To directly test the role of translational control in the cellular radioresponse, Lü et al. used microarray analysis to identify changes in polysome-bound mRNA after irradiation of the U87 glioblastoma (GBM) cell line.¹¹ For comparison to the traditional approach of investigating radiation-induced gene expression, microarray analysis was also performed under the same treatment conditions using total cellular RNA. Comparison of the radiation-induced changes in polysome-bound mRNA (i.e., the translatome) to those detected when analyzing total cellular mRNA (i.e., the transcriptome) revealed that radiation affected substantially more genes (approximately 10-fold) in the translatome than in the transcriptome, with no genes in common between the data sets. For genes in the radiation-induced translatome, northern blot analyses showed corresponding changes in their monosome/polysome distributions, consistent with radiation affecting the translation of pre-existing mRNAs. Moreover, in contrast to the transcriptome, the radiation-induced changes in the translatome were strongly correlated with changes in protein. Thus, this initial study suggested that radiation does modify gene expression and does so at the level of translational control.

In general, to identify genes subject to translational control it is necessary to perform parallel analyses of total cellular RNA (or cytoplasmic RNA) and polysome-bound RNA; those genes whose levels were affected in the polysome-bound RNA and not in the total RNA can then be considered under translational control. With respect to radiation, there was little to no overlap in the genes affected in the analysis of total RNA versus polysome-bound RNA,^11,20 suggesting that transcriptional and translational regulation are not coordinated events after irradiation. However, transcription-based events can ultimately influence gene translation, which can complicate the investigations of translational control. To reduce the potential for such secondary effects on translation, polysome analysis is performed at relatively short time points of 6h or less after treatment.²¹

In addition to polysome analysis, changes in mRNA translation can be determined using ribosome profiling, which involves the deep sequencing of ribosome-protected mRNA fragments. In essence, this method determines the number of ribosomes associated with an mRNA as well as the ribosome position. In contrast to polysome profiling, ribosome profiling can be used to distinguish between initiation and elongation events and is less sensitive to mRNA stability issues due to the degradation of unprotected mRNA and short length of the remaining mRNA fragment. Polysome profiling involves separating ribosome-associated mRNAs on a sucrose gradient based on the number of ribosomes attached with ribosome association indicating actively translated messages, followed by either microarray or RNA-Seq analyses.^22,23 This method allows monitoring of the entire transcript, which can provide information about translation rates of alternative transcripts. Polysome profiling also distinguishes a decrease in the number of ribosomes along all copies of an mRNA from a decrease in ribosomes bound to a subpopulation of mRNAs, allowing it to better predict changes in protein synthesis.^22,24

If the radiation-induced translational control of gene expression plays a role in determining radiosensitivity/resistance, then in contrast to changes in the transcriptome, some degree of commonly affected genes among cell lines would be expected. This was the case for GBM cell lines; about 30% of the genes in the radiation-induced translatomes were in common among the 3 lines.¹¹ A more extensive comparison of radiation-induced translatomes was performed by Kumaraswamy et al. in which 18 human cell lines were analyzed, including 4 tumor histologies typically treated with radiotherapy (GBM, pancreatic carcinoma, breast carcinoma, and lung carcinoma) as well as 4 normal human cell lines.²⁵ Hierarchical cluster analysis showed that glioma and pancreatic cell lines clustered according to tumor type with breast and lung tumor cell lines being considerably more heterogeneous and clustering together. The normal cell lines (1 skin fibroblast, 2 lung fibroblasts and 1 mammary epithelial) displayed extensive homogeneity forming a definitive cluster distinct from the tumor cells. Although there were clearly cell line-specific effects, each histology contained a significant number of commonly affected genes.

Tumor type selectivity in radiation-induced gene expression could have a number of implications. If gene expression influences tumor cell radioresistance, then the optimal preclinical development of targets for radiosensitizers should take into consideration tumor type. In addition, independent of whether the induced changes in gene expression directly contribute to radiosensitivity, its tissue type dependency may provide a source of biomarkers indicative of radiation exposure during cancer treatment as well as under other environmental circumstances. Finally, as shown,²⁵ there appears to be a distinct difference between cancer and normal cells in terms of radiation-induced gene expression. The distinctly different gene sets and overall greater degree of homogeneity among the normal cell lines suggest that the translational control of gene expression may provide the basis for a long sought goal in radiation oncology: the identification of tumor specific targets for radiosensitizers.

Glioblastoma-stem-like cells

Whereas initial studies showed that the translational control of gene expression occurs after irradiation, how this process specifically participates in cellular radioresponse remained to be defined. Toward this end, we focused on a panel of glioblastoma stem-like cell (GSC) lines. The rationale for investigating this cell type was 2-fold: radiotherapy remains a primary treatment modality for GBM and, whereas the biology of standard glioma cell lines has little in common with GBM in situ, GSCs are considered to be a subpopulation of tumor cells critical to GBM radioresistance.^26,27 To identify the genes subject to radiation-induced translational control in GSCs, microarray analysis of RNA bound to polysomes was performed on 3 GSC lines (NSC11, 0923 and GBMJ1) after exposure to the clinically relevant dose of 2Gy. As shown in Fig. 1, absorption profiles of sucrose-gradients generated from cells collected at 1h and 6h after exposure to 2Gy were similar to that of untreated cells, illustrating that radiation does not significantly alter global translation. Radiation, however, did affect (increase or decrease) the polysome association of greater than 1000 specific transcripts in each of the GSC lines, indicating that radiation selectively regulates gene translation in GSCs. As for results from established cell lines, comparison of the radiation-induced translatomes (polysome-bound RNA) and transcriptomes (total cellular RNA) from each of the GSCs revealed few commonly affected genes, consistent with translational control as the primary mode through which radiation modifies gene expression.

Figure 1. Influence of radiation on global translation. Representative polysome profiles of NSC11 cells after mock irradiation (Control) and 1h and 6h after 2Gy. Cells were plated in monolayer, irradiated and polysomes isolated as reported previously.²⁰ Graphs shown are representative of 3 biological replicates. Translational efficiency (TE), a measure of the relative amount of mRNA associated with polysomes and thus global translation, was estimated as the proportion of the integrated area under the curve containing polysomes relative to the entire area under the curve.⁶⁵ TE values are averages of 3 biological replicates and are not significantly different using Student's t-test, p ≤ 0.05.

To begin to address the functional consequences of the translational control of gene expression, the GSC radiation-induced translatomes were subjected to network, pathway and function analyses using IPA and GSEA. First, as an approach to evaluating the potential for translational control to identify tumor specific processes regulating radiosensitivity, the common GSC translatome was compare with those generated from the normal astrocytes and pericytes after exposure to 2Gy. As shown by the Venn diagram in Fig. 2A, although the radiation-induced translatomes from the normal CNS cells had a number of genes in common with the GSCs, the majority of the genes affected by radiation in normal CNS cells were unique to the specific cell type. Hierarchical clustering of normalized enrichment scores of GO terms showed that the normal CNS lines were more similar to each other than to individual GSCs (Fig. 2B). Thus, for these CNS phenotypes, the data presented in Fig. 2 are consistent with tumor specific aspects in radiation-induced translational control of gene expression.

Figure 2. Comparison of radiation-induced translatomes from normal CNS cells and GSCs. A) Venn diagram comparing the overlap in genes identified by microarray to be up- (green) or down- (red) regulated in the common GSC radiation-induced translatome²⁰ and the radiation-induced translatomes of Human astrocytes and pericytes (ScienCell) 1 and 6h after 2Gy irradiation. B) Unsupervised hierarchical cluster analysis of normalized enrichment scores of GO terms changed in indicated cell lines at 6h after 2Gy (green: enrichment of GO term after radiation; red: enrichment of GO term in control).

The GSC translatomes revealed cell line specific changes as well as a significant number of genes whose translation was affected in more than one of the GSC lines, referred to as the common translatome. A number of pathways detected in the common radiation-induced translatome were associated with the DNA damage response, a critical determinant of radiosensitivity. Enhancing the translation of genes involved in DNA repair and cell cycle checkpoints suggests a critical role for translational control in protecting against radiation-induced cell death. Support for such a role for the translational control of gene expression in the DNA damage response comes from studies using the mTOR inhibitor INK128, which significantly reduces overall translation.²⁸ Treatment of GSCs with INK128 inhibits the repair of radiation-induced DNA double strand breaks and enhances radiation-induced cell death.²⁰ Of note, in these experiments INK128 was added after irradiation, consistent with inhibiting a radiation-induced process such as translational control. INK128 was also shown to inhibit DNA repair and enhance radiosensitivity when added after irradiation to a panel of pancreatic tumor cell lines. Moreover, exposure of the PSN1 pancreatic cell line to INK128 dramatically reduced the number of genes in the radiation-induced translatome, including those involved in the DNA damage response.²⁹ The results gathered from a number of tumor cell lines thus suggest that the translational control of gene expression plays a role in DNA repair and cell survival. More specifically, these results suggest that mTOR can serve as a clinically relevant target for tumor radiosensitization.

Further analyses of the common radiation-induced GSC translatome revealed the activation of additional processes not traditionally associated with radioresponse. Upstream signaling analysis identified 23 genes linked to the activation of mTOR and eIF4E. Given that eIF4E is a critical component of eIF4F and that mTOR regulates the availability of eIF4E, this analysis suggested that radiation enhances cap-dependent translation. Testing this scenario, radiation was found to increase eIF4F-cap complex formation in each of the GSCs evaluated, consistent with an increase in cap-dependent translation, which was eliminated by INK128. Because the majority of mRNA translation occurs in a cap-dependent manner, its modification would seem a reasonable event in radiation-induced translational control. However, whereas cap-complex formation was increased in GSC lines after irradiation, it has been reported to be reduced in a normal fibroblast cell line³⁰ and unaffected in a breast tumor cell line.³¹ Thus, the significance of elevated cap-complex formation in translational control after irradiation is unclear. Cap-complex formation is one of the last and a rate-limiting event in gene translation and can be affected by each of the upstream components of post-transcriptional regulatory process.³² Because mTOR kinase inhibitors reduce translation in general,³³ it may be that a basal level of translation is required for radiation-induced translational control of gene expression and that variations in cap-binding are more likely to influence the specific genes affected.

The common radiation-induced GSC translatome was also enriched in GO terms associated with Golgi function. Of interest, Farber-Katz et al. reported that DNA damaging agents, including radiation, induce a change in Golgi morphology reflected by a change in the distribution of the protein GM130 from perinuclear to more fragmented and cytoplasmic, referred to as Golgi dispersal.³⁴ As suggested by the GSC translatome analyses and consistent with data generated from a variety of established cell lines, Golgi dispersal was detected after irradiation of GSCs. However, the initial report by Farber-Katz et al. indicated that radiation-induced Golgi dispersal was the result of an interaction of DNA-PK with specific Golgi proteins.³⁴ In GSCs the radiation-induced Golgi dispersal was prevented by the mTOR inhibitor INK128, suggesting a requirement for cap-dependent translation and consistent with the radiation-induced translational control of gene expression. Whether radiation-induced Golgi dispersal is the result of more than one process that is cell type specific or the 2 apparently different mechanisms are related in some manner remains to be determined. Regardless of the mechanisms involved, the role of Golgi dispersal in cell survival after irradiation remains to be clearly defined.

In addition to common components of GSC radioresponse, analyses of the individual radiation-induced translatomes identified cell line-specific responses with the most apparent being mitochondrial structure and function. Specifically, translatome analyses predicted that mitochondria function after irradiation would increase in NSC11, decrease in 0923 cells and not be affected in GBMJ1 cells. These predicted changes after irradiation were validated according to analysis of mitochondria mass and mitochondrial DNA content after GSC irradiation. Mitochondrial parameters after irradiation have been the subject of a number of studies and appear to be cell line dependent with increases, decreases and no effects being reported.^35-38 Whereas most studies address the radiation-induced damage to nuclear DNA, it can also induce detectable damage in mitochondrial DNA (mtDNA),^38,39 resulting in oxidative stress and altering mitochondrial function.^40,41 An increase in mitochondrial content, which can result from an increase in mitochondrial volume or number, has been suggested to be compensatory mechanism aimed at maintaining mitochondrial function.^42-44 Although the role of mitochondrial function in radioresponse has not been defined, the results of polysome profiling suggest that it is at least partially regulated by translational control and is cell line-dependent.

Mechanisms

Data generated to date indicate that the radiation-induced translational control of gene expression is a fundamental component of cellular radioresponse. Although further investigation is clearly required, it appears that the mechanisms mediating translational control after irradiation are consistent with the post-transcriptional operon model initially put forth by Keene and Tenenbaum.^45,46 This model proposes that post-transcriptional regulation of gene expression is mediated by mRNA binding proteins (RBPs) that bind to sequence specific elements in common among functionally related mRNAs. More specifically, RBPs act in a combinatorial manner to regulate the splicing, nuclear export, stability, and translation initiation of functionally related transcripts.^45,46 Of particular relevance to the radiation-induced translatome, given that there is little in common with the radiation-induced transcriptome,^11,20 is that many RBPs regulate translation at the initiation step by promoting or inhibiting proper cap-binding complex formation of a given transcript, which can involve either directly or indirectly recruiting additional proteins.^47,48 and/or by influencing transcript stability.^49-51

A role for RBPs in radiation-induced translational control is directly supported by studies linking their transcript interactions to ATM and DNA-PK, apical kinases in the radiation-induced DNA damage response. Along these lines, a number of reports have focused on Human antigen R (HuR), which binds to adenylate-uridylate-rich elements (AREs) in the 3’ UTRs of transcripts and regulates several aspects of post-transcriptional gene expression, including mRNA stability, translation, and splicing.⁵² Activation of ATM in response to DNA double strand breaks leads to the phosphorylation of CHK2 (among other proteins), which then phosphorylates HuR.⁵³ Using a ribonucleoprotein immunoprecipitation assay followed by microarray analysis, Mazan-Mamczarz et al. showed that irradiation of lymphocytes altered the mRNAs bound to HuR 6h after 1Gy.⁵⁴ Transcripts that increased binding with HuR after irradiation were more actively translated; transcripts that decreased binding to HuR were more translationally suppressed. Importantly, these changes in translation were not observed in ATM-null cells. Masuda et al. followed a similar approach using HCT116 colorectal adenocarcinoma cells and found that HuR dissociated from most of its transcripts 30 minutes after 10 Gy. Linking this process to ATM, in CHK2 null HCT116 cells HuR was not phosphorylated after irradiation and remained associated with most of its transcripts.⁵⁵ As observed in lymphocytes, HuR modified radiation-induced gene expression at the translational level. Using clonogenic survival analysis, Masuda et al. went on to show that dissociation of HuR from its mRNA clients enhanced radiation-induced cell killing, illustrating the functional relevance of this pathway. Thus, data suggest that one aspect to radiation-induced translational control involves the ATM-mediated activation of HuR.

In addition to HuR, ATM and/or DNA-PK have been reported to phosphorylate the RBPs FUS,^56-58 hnRNP C and hnRNP A1.^59,60 Interestingly, data from GSCs shows that a number of RBPs themselves, including members of the hnRNP family, are translationally regulated in response to radiation, adding another layer of post-transcriptional regulation to radiation-induced gene expression.²⁰ As an approach to defining the RBPs participating in radiation-induced translational control, we are currently analyzing the 3′ and 5′ untranslated regions (UTRs) of the radiation-induced translatome of GSCs in an attempt to identify enriched sequence motifs. A study utilizing stable isotope labeling by amino acids in cell culture (SILAC) identified hundreds of ATM targets,⁶¹ with one network associated with RNA post-transcriptional modification⁶¹; whereas not as thoroughly investigated, DNA-PK also has a number of known and likely unidentified substrates.⁶² Given the critical role of ATM and DNA-PK in regulating overall cellular radioresponse, it would seem that they could influence additional RBPs and other proteins involved in translation to play a significant role in the radiation-induced post-transcriptional control of gene expression. However, it is likely that other signaling molecules are involved, which will then influence the activity of additional RBPs. For example, as compare with their wild type proteins, p53 and mutant K-Ras have been reported to modify translational gene expression profiles.^63,64 Because both proteins participate in signaling pathways that influence cellular radioresponse, they may also play a role in mediating radiation-induced translational control. The putative involvement of multiple signaling molecules along with the combinatorial actions of RBPs would seem to account for the number of genes subject to radiation-induced translational control as well as the differences between tumor and normal cells. Clearly, considerable research is required to define the molecules and processes mediating radiation-induced translational control. However, initial studies suggest that understanding how radiation impacts gene expression will offer novel insight into the cellular radioresponse and may provide opportunities to improve its therapeutic efficacy.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Funding

Financial support has been provided by the Division of Basic Sciences, Intramural Program, National Cancer Institute (Z1ABC011372, Z1ABC011373)