Analysis of lncRNA-miRNA-mRNA expression pattern in heart tissue after total body radiation in a mouse model

Six- to 8-week old female C57BL/6 J mice were given total-body irradiation (TBI) with X-rays using the Small Animal Radiation Research Platform (SARRP Xstrahl Ltd.). Mice were placed in plastic containers and exposed to a single surface dose of 1, 2, 4, 8, or 12 Gy at a dose rate of 1.05 Gy/min. Control mice (0 Gy) were placed in the same type of plastic container and sham irradiated. Three animals per dose were included in the study. Hearts of irradiated and control animals were harvested 48 h after TBI. Organs were snap frozen in liquid nitrogen and stored at − 80 °C until processed for RNA isolation. All animal experiments were performed at the Department of Pathology at New York University (NYU) Langone Medical Center under an approved IACUC protocol as part of a collaborative study.

Samples were bathed in liquid nitrogen and pulverized into a fine powder using a mortar and pestle. Approximately 100 µg of powdered sample was lysed with 700 µl of QIAzol lysis buffer (Cat # 79306, QIAGEN) and homogenized by passing the solution through QIAshredder spin columns (Cat # 79654, QIAGEN). RNA isolation was performed using standard miRNeasy mini kit (Cat # 217004, QIAGEN) according to the manufacturer’s protocol. Quality and quantity of the RNA samples were assessed using a DeNovix DS-11 nanodrop spectrophotometer (DeNovix, DE, US) and Agilent Bioanalyzer with the RNA6000 Nano Lab Chip (Agilent Technologies, Santa Clara, CA).

Microarray analysis was performed for sham animals (0 Gy) and 1 Gy, 2 Gy, 4 Gy, 8 Gy, and 12 Gy irradiated animals. Quality assessments and microarray experiments were completed as previously reported [46]. Samples were hybridized to Agilent Mouse GE 8 × 60 K v2 arrays for mRNA expression analysis and to Agilent Mouse miRNA 8 × 60 K v21.0 arrays (Design ID 070155) for miRNA expression analysis. Slides were washed and scanned on an Agilent SureScan Microarray Scanner. Expression values were extracted using Agilent Feature Extraction software and data were analyzed with GeneSpring GX software (Agilent Technologies).

1000 ng of total RNA was reverse transcribed using RT2 First Strand Synthesis kit (Qiagen, US). Individual RT-qPCR reactions using RT2 qPCR primer assays and RT2 SYBR Green qPCR Master Mix (QIAGEN, US) were performed for the following lncRNAs: Trp53cor1 (Assay ID No. LPM12776A), Dino [47] (FP- GCAATGGTGTGCCTGACTAT; RP- TTCTGGCTTCCCAGAG), Pvt1 (LPM16140A) and Rplp0 (assay ID no. PPM03561B) in the 48 h mouse heart tissue samples. Relative expression was calculated as: 2^−dCt where dCt = Ct [test gene] − Ct [Rplp0] [44]. Following qPCR Primer Assay were used for mRNA validation PPM03371A- Ckap-2, PPM05273A-Alas2, PPM04436C-Gdf15, PPM02901B-Cdkn1a, PPM03154A-Cx3cr1, and PPM04813A-Aplnr. Rplp0 was used as the normalizing control in both lncRNA and mRNA PCR assays.

200 ng of total RNA was used for first-strand cDNA synthesis reactions using miRCURY LNA RT Kit (Cat. No. 339340) according to the manufacturers protocol. Reverse-transcription reaction was done at 42° for 60 min, followed by an inactivation step at 95° for 5 min. Quantitative Real-Time PCR was done using individual miRCURY LNA miRNA PCR Assays (Cat. No. 339306) for the following primers (mmu-miR-103a-3p, mmu-miR-149 3p, mmu-miR-211-3p, mmu-miR-3960, mmu-miR-6538, mmu-miR-7118-5p, mmu-miR-8101) to detect differential expression in irradiated vs. control samples. Real time PCR reactions were performed using Applied Biosystems Quant Studio Real-Time PCR machine. PCR steps included initial heat activation at 95° for 2 min followed by two step cycling: Denaturation at 95° for 10 s followed by combined annealing/extension at 56° for 60 s for 40 cycles. A melt curve analysis was performed to ensure the specificity of the corresponding RT-qPCR reactions. Fold change = 2^−ddCt where ddCt = dCt (irradiated) − dCt (control); dCt = Ct (gene) − Ct (endo control: UniSp6); and Ct is the threshold cycle number. All assays were performed in triplicates. Statistical significance was calculated using student’s unpaired t-test.

Analysis of mRNA and miRNA data was performed using R statistical software and the Bioconductor Linear Model for Microarray Analysis (LIMMA) package in R [48]. Background correction and normalization were performed in R using the normal-exponential correction method and quantile normalization between arrays [49]. Only probes with intensities above background on at least one array were kept in the dataset for analysis. Transcripts with multiple probes were averaged such that the final set reflected best estimates of transcript level expression. A linear model was fit to each probe to assess differential expression for pair-wise dose comparisons within the heart-tissue samples. This method employed an empirical Bayes smoothing approach that results in more stable model estimates by using information on variance from the whole probe set, despite the small number of arrays. Models were developed for each of the pair-wise comparisons between each dose (1, 2, 4, 8, and 12 Gy) and the control probes (0 Gy), and resulting probes were filtered using log₂ fold change and adjusted p-value thresholds (|log₂FC| > 1, adjusted p-value < 0.05) [50]. Additionally, a nested interaction model was fit for each probe to examine dose within tissue as a linear (continuous) trend. Each model yielded main effects for the heart tissue and dose within the heart tissue. Probes were filtered using the nested dose coefficients with log fold change and adjusted p-value thresholds (|log₂FC| > 1, adjusted p-value < 0.05). Finally, gene ontology analysis was utilized to identify affected pathways from the differentially expressed probes.

To identify potential interactions, paired analysis was conducted to evaluate correlative relationships between pairs of differentially expressed mRNA and miRNA probes. mRNA and miRNA probes were paired using shared target transcript Ensembl IDs [51]. Probes that could not be mapped or paired were excluded. Transcripts for miRNA probes were identified using an Agilent microarray gene dataset and the TargetScan database; transcripts for mRNA probes were identified using an Agilent microarray gene dataset [52]. Transcript-miRNA pairs with a TargetScan context++ score above − 1 were excluded. Probe pairs with differentially expressed miRNA and mRNA probes were identified within the heart tissue for continuous dose contrast models. Pearson correlation coefficients of miRNA and mRNA expression across all experiments were calculated and plotted for the differentially expressed probe pairs.

Both core and comparison analyses were performed in IPA (QIAGEN Inc., https://​www.​qiagenbioinforma​tics.​com/​products/​ingenuitypathway​-analysis). Pathways and function terms that satisfied an absolute z-score > 2 and p-value < 0.01 were predicted to be altered based on the gene expression data.

Microarray analysis performed on all mouse heart samples revealed 2041 differentially expressed genes (|log2FC| > 1; p-value < 0.05) that distinguished unirradiated control samples from samples of at least one dose of TBI mice. Overall, most genes had relatively low to no basal expression in control samples and showed increased expression levels after TBI; however, a cluster of genes showed relatively pronounced high expression in control samples that decreased to low expression after irradiation (Fig. 1A). For each dose, more genes were differentially upregulated than downregulated (Fig. 1B). Across all doses, 99 genes were commonly expressed in response to radiation and 128, 55, 390, 322, and 316 genes were expressed exclusively after 1, 2, 4, 8, and 12 Gy of TBI, respectively (Fig. 1C). Additional file 2: Table S1 lists fold changes and p-values of all differentially expressed genes by dose. While there was no systemic dose–response in terms of the number of genes expressed, we did observe more differentially expressed genes in the higher doses (4, 8, 12 Gy) than in the lower doses (1, 2 Gy). When the dose response of each gene was analyzed by fitting a linear model to each probe, 596 probes were found to have significant dose-responsive up- or down-regulation across all doses; here we present the 20 most upregulated and downregulated genes (Additional file 3: Table S2). Cdkn1a, Ckap2, and Gdf15 were among the top 20 probes with the strongest upward linear trend, and Alas2, Aplnr, and Cx3cr1 were among the top 20 probes with the strongest downward linear trend (Fig. 1D). All six genes have previously been reported in the context of radiation or DNA damage response and fall into three main biological roles: cell cycle arrest, hemoglobin metabolism, and inflammatory response (Table 1). The prior relevant literature on Cdkn1a [53, 54], Gdf15 [55, 56], Ckap2 [57, 58], Alas2 [59, 60], Aplnr [61, 62] and Cx3cr1 [63, 64] are listed in Table 1. We validated the expression of the most significantly upregulated and downregulated mRNAs in our data using RT-qPCR (Additional file 1: Figure S1).

To understand the response of heart-based lncRNAs to TBI, we filtered whole genome microarray data to include only probes that correspond to transcripts of lncRNAs. Of the 87 lncRNA transcripts in the microarray data that passed the background intensity cutoff in at least one condition, 46 were differentially expressed in response to radiation, irrespective of the TBI dose (Fig. 2A). Most lncRNAs showed relatively low expression in unirradiated control samples with increased expression after radiation. More probes showed upregulation than downregulation in all doses except 2 Gy, which had 4 downregulated lncRNAs and 3 upregulated lncRNAs (Fig. 2A, B). Two lncRNAs were significantly altered at all doses after radiation, including: chr10:69819062-69871640_F and chr17:29183003-29217681_R (Trp53cor1-up) (Additional file 4: Table S3) (Fig. 2C). Additional lncRNAs were altered only at specific doses, with 1, 1, 7, 7, and 10 lncRNAs expressed exclusively in 1, 2, 4, 8, and 12 Gy, respectively (Fig. 2C). Additional file 4: Table S3 lists the fold changes and p-values for the differentially expressed lncRNAs at each dose. Twenty probes showed significant linear upward or downward trends as the dose of TBI increased, demonstrating a linear dose response (Additional file 5: Table S4). Abhd11os, Trp53cor1, Pvt1, and Kalrn were among the most significant annotated lncRNAs that became upregulated as radiation dose increased, while the linc-RAM (Malrn) transcript had the most significant dose-responsive downregulation (Fig. 2D). Trp53cor1 was the most sensitive to radiation, showing significant increase in the relative intensity in comparison to the unirradiated control even after 1 Gy of TBI. The basal level expression of Trp53cor1 lncRNA expression was below detection threshold levels in unirradiated heart tissue. Due to this reason, we used relative intensity to describe expression of Trp53cor1 after radiation.

In contrast, basal expression of Abhd11os was at a much higher threshold across all doses, including control samples, with significantly higher expression levels after 8 and 12 Gy TBI. Pvt1 showed significance after 2 Gy and Abhd11os showed significance after 8 Gy, while Kalrn and linc-RAM showed significance only after 12 Gy TBI. We confirmed the expression of Dino, Pvt1, and Trp53cor1 in heart samples through RT-qPCR (Fig. 2E). In concordance with the microarray data, Trp53cor1 showed very low expression in control samples but significant dose-responsive upregulation after radiation. Pvt1 also showed consistent results with the microarray in terms of the dose response; however, it showed significance in expression change only after 12 Gy TBI. Damage induced noncoding lncRNA (Dino) was not present in our microarray data due to lack of a probe, but prior data from our lab led us to validate its expression in the heart via RT-qPCR. We found that Dino is also significantly expressed in the heart after every dose of TBI. Like Trp53cor1, Dino showed very low expression levels in control samples but increased significantly after radiation (Fig. 2E).

A separate whole genome microarray analysis of miRNA expression revealed 102 significantly altered miRNAs in mouse heart tissue in response to radiation. Surprisingly, the largest and most significant changes in expression occurred after 1 Gy of TBI, with 86 differentially expressed miRNAs identified at this dose (Fig. 3A). Furthermore, there were no commonly expressed miRNAs across all doses and no miRNAs significantly expressed in 4 or 8 Gy TBI samples (Fig. 3B, C). We did observe significant regulation of miRNAs at 2 and 12 Gy; however, the numbers were relatively low, with 1 and 19 miRNAs in 2 and 12 Gy, respectively. Fold changes and p-values for all significantly altered miRNAs at each dose are listed in Additional file 6: Table S5. Linear trends were fit to all miRNA probes to identify dose response across all doses and found 21 probes that showed significant upward linear trend (Additional file 7: Table S6). No miRNAs showed a significant downward linear trend. Among the top linear probes were miR-149-3p, miR-6538, miR-3960, miR-8101, miR-7118-5p, and miR-211-3p, all of which showed a statistically significant increase in expression only after 12 Gy of TBI (Fig. 3D). We validated expression of miR-149-3p, miR-3960, miR-7118, miR-6538, miR-8101, and miR-103a-3p (Additional file 8: Figure S2). miR-211-3p was below detection using RT-qPCR (data not shown).

Since we are ultimately interested in developing integrated signatures of coding and non-coding RNA response to radiation, we sought to identify potential interactions of the miRNA and mRNA signatures and their biological significance. Using IPA, we first conducted miRNA target filter analysis of the significant differentially expressed miRNAs to identify experimentally verified mRNA targets in our dataset. Canonical pathway analysis of the identified targets revealed significant activation of pathways relevant to cell cycle checkpoint activation and senescence, including p53 signaling and numerous apoptosis signaling pathways, among others (Fig. 4A). Interestingly, most pathways activated across all doses were predicted to have the highest activation after 4 Gy TBI, followed by lower activation in 8 and 12 Gy. One exception to this was the senescence pathway, which showed the lowest activation at 4 Gy. A concurrent higher activation of various apoptosis pathways after 4 Gy TBI may suggest that miRNA-mRNA pairs at this dose exhibit a different response to stress and DNA damage than both lower and higher doses at this time-point after irradiation. Interestingly, analysis of the mRNA targets of the significant miRNAs revealed the strongest activation and deactivation of pathways after 4 Gy TBI, despite a majority of differential miRNA expression after 1 Gy TBI. Since pathway analyses are based on mRNA rather than miRNA data, it is likely that this observation arises from the more pronounced mRNA expression at 4 Gy compared to 1 Gy, After pathway analysis, we identified predicted miRNA-mRNA pairs with inverse expression patterns, demonstrating their potential for inclusion in an integrated RNA marker signature to improve clinical decision making. We found three miRNAs that showed significant expression in at least one dose and had predicted targets with inverse expression patterns, each miRNA with two targets. Radiation decreased expression of both miR-128-3p and miR-122-5p relative to control, while their targets—Tgfbr1 and Wee1, and Fam117b and Slc7a11, respectively—showed a statistically significant increase in expression as radiation increased (Fig. 4B). The third miRNA, miR-18-5p, showed increased expression after radiation, most significantly after 1 Gy. Its targets, E2f1 and E2f2, showed significant down regulation across all doses.

While understanding the interactions and biological implications of the miRNAs and mRNAs is critical for developing an integrated biomarker signature, we hypothesized that because the relatively low number of differentially expressed miRNAs would limit the number of mRNAs included in the pathway analysis, we could potentially miss genes that play a significant role in pathway regulation. Therefore, we also conducted a canonical pathway analysis using all differentially expressed mRNAs, irrespective of interactions with miRNAs in our dataset. A similar overall pattern of pathway regulation exists between the target mRNAs and target/non-target mRNAs, with significant deactivation of most of the pathways involved (Fig. 5A). Several pathways related to coagulation, including both ex- and intrinsic prothrombin activation and the coagulation system pathway, were downregulated across all doses. Changes in immune-related pathways were less consistent in terms of activation or deactivation. While natural killer cell signaling was activated significantly after 4 and 12 Gy TBI, the complement system was inhibited across all doses.

Two clusters of pathways showed the most pronounced activation and deactivation. Five pathways were predicted to be activated across all doses, and all were involved in immune response or cell cycle regulation. There were 78 genes differentially expressed in these pathways, a majority of which were downregulated with respect to the control (Fig. 5B). A full list of genes can be found in Additional file 9: Table S7. In contrast to the pathway analysis of the mRNA targets, which showed the highest activation of several pathways after 4 Gy TBI, all activated pathways except natural killer cell signaling were consistently activated as the dose increased. We observed that radiation inhibits pathways relevant to xenobiotic metabolism and biosynthesis of lipids, including hormones. This is demonstrated by activation of LPS/IL-1 mediated inhibition of RXR function and inhibition of the super-pathway of melatonin degradation, among others. There were 79 genes involved in deactivation of these seven pathways, most of which were downregulated after TBI (Fig. 5C, Additional file 9: Table S7).

IPA analysis predicted significant deactivation of several metabolic pathways, such as triacylglycerol degradation and type I diabetes mellitus signaling, among others (Fig. 5A). To further understand how radiation alters metabolism, we used IPA to filter the microarray gene expression data to include only the significantly differentially expressed genes with involvement in metabolic energy production pathways, with emphasis on fatty acid oxidation (Fig. 6A). Of note, there was a significant downregulation of solute carrier 2a2 (Slc2a2) and concurrent upregulation of pyruvate dehydrogenase kinase 4 (Pdk4), especially after 4 Gy TBI. Similarly, as dose increased, we observed upregulation of glutamine synthetase (Glul), which encourages conversion of glutamate to glutamine. We also observed downregulation of glutaminase (Gls2), which converts glutamine to glutamate.

Gene expression data for fatty acid metabolism at 48 h post radiation was contradictory. We observed a downregulation of solute carrier 27a5 (Slc27a5), which is associated with fatty acid entry into the cell; an upregulation of sestrin 2, which inhibits lipid catabolism; and an upregulation of hormone sensitive lipase (Lipe) and adipose triglyceride lipase (Atgl or Pnpla2), which cleave lipid droplets to allow use of fatty acids in FAO. However, there was a downregulation of arylacetamide deacetylase (Aadac) which shares homology with Lipe and is also thought to control triglyceride levels. Additionally, acyl-CoA synthetase medium chain family members 1 (Acsm1) was downregulated, while acyl-CoA thioesterases 1 (Acot1) and acyl-coA dehydrogenase 10 (Acad10) were notably upregulated as radiation doses increased.

The observed early changes in gene expression led to significant dysregulation of pathways commonly associated with radiation exposure, including cell cycle arrest, apoptosis, and senescence. The most significantly altered genes, Cdkn1a and Ckap2, are capable of inducing cell cycle arrest or apoptosis under stress conditions [70, 71]. The observed upregulation of Gdf15 and downregulation of Aplnr are associated with stress induced senescence and have previously been reported as biomarkers of radiation exposure [56, 72, 73]. The inversely correlated expression of miRNAs and their mRNA targets that are associated with these pathways provides insight into the potential mechanisms of acute effects of TBI and targets to mitigate acute and delayed effects on heart tissue. Prior research demonstrated miR-128 negatively regulated Wee1 and Tgfbr1, genes involved in mitotic inhibition [74‐76]. Additionally, miR-122 is known to increase radiation sensitivity and targets Slca11 and FAM117b [77]. While little is known about the function of Fam117b, Slc7a11 downregulation is associated with RT-induced ferroptosis in tumor cells [78, 79]. The miR-17-92 cluster represses E2F1-3 to regulate cell proliferation and apoptosis, and includes miR-18a [80]. A negative feedback loop has been observed within this cluster, as E2F1 upregulates miR-17-92 which causes increased repression of the genes E2F1-3 [81]. Mechanistic studies are required to confirm the miRNA-mRNA interdependent functions in our data. Similarly, further research into lncRNA-mRNA functions are also needed. Acute lncRNA Dino overexpression caused an increase in Cdkn1a expression in cervical cancer cells [82]. Our data suggests an interplay between both lncRNA Dino and mRNA Cdkn1a expression.

Anemia and decreased hemoglobin levels are a known side effect of RT [83]. We observed decreased expression levels of genes relevant to hemoglobin synthesis. Alas2, the rate limiting enzyme in heme synthesis, became increasingly downregulated as radiation dose increased. Similar downregulation of Alas2 was recently cited as a potential predictive marker for radiation induced hematological toxicity in cancer patients [60]. Other genes associated with hemoglobin synthesis, including Hbb-bt, Hbb-b1, and Hba-a1, were also significantly, linearly downregulated (Additional file 3: Table S2).

Our pathway analysis also indicated inflammatory pathways are downregulated at higher dose levels at 48 h post-radiation. Of note, TBI induced significant downregulation of Cx3cr1, which is known to induce recruitment of immune cells and an inflammatory response in smooth muscle and endothelial cells [84]. Short-term data contrasts long-term in vivo data that showed upregulation of inflammatory markers in the heart of a mouse model 40 days post-irradiation [17]. Additional studies that include intermediate timepoints are needed to clarify when the heart transitions from an anti-inflammatory to a pro-inflammatory response after radiation to enable improved treatment options.

Consistent with previous research [85], our findings indicated inhibition of extrinsic and intrinsic prothrombic pathways that couple with inhibition of coagulation pathway after radiation exposure. Radiation is known to increase likelihood of coagulopathy, which can lead to death when untreated [86]. We also observed downregulation of hemoglobin subunit beta (Hbb) and hemoglobin subunit alpha (Hba) (Additional file 3: Table S2). In combination with anemia, failure to clot produces hemostatic dysfunction and potential death, though the pathogenesis is poorly understood [87]. Our research highlights these alterations in gene expression to provide insight into potential mechanisms of and therapeutic targets for acute radiation syndrome (ARS).

Retinoid X receptor and liver X receptor (RXR/LXR) activation is associated with protection against heart failure due to their role in improving glucose tolerance, decreasing lipid accumulation, and decreasing inflammation [88]. RXR signaling has previously been shown to increase estradiol synthesis from pregnenolone [89]. We observed that the estrogen and pregnenolone biosynthesis pathways are also inhibited. Inhibition of RXR/LXR and its downstream pathways coupled with inhibition of triacylglycerol degradation may indicate a deleterious increase in lipid accumulation within the heart.

Previously, we reported the concomitant differential expression of p53-related lncRNAs such as Pvt1, Dino, Trp3cor1 after TBI in a mouse model [44]. In the current study, we observed significant alterations in abhydrolase domain containing 11, opposite strand (Abhd11os), Pvt1, Trp3cor1, Kalrn, linc-RNA activator of myogenesis (Linc-RAM), lncRNA chr10:69819062-69871640_F and Dino. Increased expression of Abhd11os has previously been shown to decrease lesion size in a Huntington’s disease mouse model and the authors reported its crucial roles in neurodegenerative diseases [90], but the exact mechanism is still unclear. Linc-RAM encourages adult skeletal muscle stem cells to differentiate into skeletal muscle through myogenic differentiation (MyoD), which aids in muscle repair after injury [91, 92]. While Linc-RAM has not been directly associated with cardiomyocytes, MyoD-null dystrophin-null transgenic mice develop severe cardiomyopathy [93]. Pvt1 has been associated with radiation resistance in cancer cells and cardiac hypertrophy in cardiomyocytes [94‐96]. Previous studies showed that Pvt1 binds Cdkn1a and miR-149-3p to suppress their activity in primary chondrocytes and Burkitt lymphoma Rajit cells, respectively [97, 98]. In contrast, Gdf15 is a downstream target of Pvt1 and was positively regulated by the lncRNA in hepatocellular carcinoma cells [99]. Our data indicates an increase in expression of Pvt1, miR-149-3p, Cdkn1a and Gdf15 by 12 Gy, suggesting that the interactions between these noncoding and coding RNA must be further elucidated in normal tissue. Further, the clinical relevance of lncRNA chr10:69819062-69871640_F has yet to reported. With the observed significant expression changes across all doses of radiation, further studies are crucial to inform more conclusions about the role of this lncRNA in the heart’s response to radiation.

Prior research demonstrated the integral role that miRNAs play in cardiac fibrosis and proliferation as well as response to radiation injury [100, 101]. We therefore anticipated changes in miRNA expression in heart tissue after TBI. Surprisingly, and possibly due to the early time point and stringent statistical analysis, we only observed significant upregulation of miRNAs at 1 Gy and 12 Gy. Among the up-regulated miRNAs, miR-149-3p has previously been implicated in multiple functions, including cell migration repression and metabolic modifications in A549 cells, a non-small cell lung carcinoma (NSCLC) model [102]. Additionally, miR-211 has been demonstrated to decrease cell proliferation and metastasis in vitro in breast and renal cell carcinoma models [103, 104], while miR-3960 has been implicated in calcification, decreased elasticity, and cardiac dysfunction in vascular smooth muscle cells of male C57BL/6 mouse aortas [105]. Arterial calcification and valvular, ventricular, and diastolic dysfunction are well-known complications of RIHD disease [106]. While the functions of certain miRNAs are not well understood, previous studies indicate that miR-8101 and miR-6538 are associated with heart failure [107, 108].

Upregulation of miR-149-3p is associated with inhibition of glucose metabolism. Its role as a therapeutic target to protect against diet-induced obesity and metabolic dysfunctions was shown previously in both colorectal cancer patients, tumors taken from colorectal cancer patients and male C57BL/6 J [109, 110]. In general, the observed changes in metabolism-related gene expression suggest that fatty acids are not being used for catabolism (Fig. 6B). Increasing doses of radiation appear to inhibit glucose oxidation through increased expression of Pdk4, which uncouples glycolysis from oxidative phosphorylation by blocking pyruvate dehydrogenase (Pdh), and decreased expression of Acsm1 [111, 112]. However, we also observed a significant downregulation of the transporter Slc27a5, which would inhibit entry of fatty acids into the cell for anabolic or catabolic use. This could be a fatal side effect of IR exposure because the adult heart relies on fatty acid oxidation as its main source of energy production [113]. Since Lipe and Pnpla2 were upregulated, the heart may be relying on internal stores of triacylglycerol to produce energy. Aside from Acsm1, no overall changes in acyl-CoA synthetases were observed; these enzymes combine fatty acids with Coenzyme A for use in FAO or lipogenesis. We observed upregulation of Acot1 which plays contradictory roles in FAO as it can separate long chain fatty acids from coenzyme A to decrease the available substrate pool [114]. However, prior research indicates Acot1 also increases FAO through activation of peroxisome proliferator-activated receptor α (Ppar-α), which upregulates Acad variants [115‐117]. We only observed an upregulation of Acad10 and there was no significant change to Ppar-α or other Acad variants at 48 h after radiation. Furthermore, there was no change in expression of other genes within the FAO pathway. This inhibition of glucose oxidation paired with an apparent reliance on FAO and an increase in free fatty acids matches what is seen in some forms of end stage heart failure [118, 119].

A recent study from our lab reported survival predictive signatures inherent to heart, lung and liver in TBI Gottingen minipigs [120]. Increased expression of Pdk4 was significantly upregulated in the hearts of non-surviving (survived < 7 days post-TBI) minipigs. We also observed increased expression of Pdk4 at 4–12 Gy radiation. As previously mentioned, the upregulation of Pdk4 is associated with decreased glucose oxidation and potential failure of energy production. Hearts from these non-surviving minipigs showed a decrease in the Apelin signaling pathway when compared to survivors and sham animals. In the present study, we also observed inhibition of the Apelin signaling pathway starting at 2 Gy. Apelin signaling is important for endothelial cell proliferation and migration and has been shown to inhibit TGF-β induced cardiac fibrosis and senescence [121]. Conversely, downregulation of Apelin has been linked to heart failure and ventricular dysfunction [122]. Both these findings implicate centrality of the endothelial cell damage in radiation induced heart injury. Finding consistency between mouse and minipig studies indicates conservation of the radiation response across species, suggesting that rescuing the function of this pathway may prevent RIHD in humans as well.

Finding normal tissue injury markers in a less invasive way is warranted for clinical applications. From this perspective, we looked at the commonality of gene and lncRNA expression changes between heart tissue of the current study and mouse whole blood after TBI from a study previously published by our group [44]. Interestingly, we detected few genes in common between both mouse heart tissue and whole blood. We focused on genes altered at 48 h after 8 Gy in whole blood and compared these alterations to changes found in our current study of mouse heart tissue. Significantly altered genes included Cdkn1a, Pmaip1, H2Aa, H2Ba1, Cx3cr1, Snca, and Gm9992 (Additional file 10: Table S8). Another previous study from our lab indicated that Pvt1 was significantly expressed in whole blood as early as 16 h after at least 2 Gy TBI, and was sustained until 48 h after TBI [44]. Similarly, at 48 h post-TBI, Trp53cor1 and Dino were significantly upregulated in the whole blood of 12 Gy and 8 Gy TBI mice, respectively. We observed similar expression patterns of these lncRNAs in heart samples after 48 h.

We also compared the heart data to a previously reported in vitro study of gene expression changes in human coronary artery endothelial cells (HCAEC) at 24 h after 10 Gy of single dose radiation [42]. The genes Cdkn1a, growth differentiation factor 15 (Gdf15), and DNA damage-inducible transcript 4 (Ddit4/Redd1) have previously been reported as radiation markers [72, 123, 124]. They showed concomitant upregulation in mouse heart tissue (Additional file 10: Table S8) and in HCAEC [42]. Additionally, we noted the upregulation of hypoxia-inducible factor 3a (Hif3a) and insulin-like growth factor 1 (Igf1) in mouse heart tissue, whereas these two genes showed significant downregulation in HCAEC in vitro study. In rat cardiomyocytes, Hif3a silencing led to increased cell viability and decreased necrosis after hypoxia challenge, suggesting decreased Hif3a expression is cardioprotective [125]. In a population study of elderly individuals, decreased serum Igf1 expression was shown to be a risk factor for mortality after ischemic heart disease [126]. These regulation inconsistencies may stem from differences between models, time points, and radiation dose rate between in vivo and in vitro experiments.

With the use of thoracic RT and the continued risk of a large-scale nuclear exposure incident inadvertently causing potential damage to the heart, understanding the effects of IR exposure on critical organs such as the heart will improve patient outcomes. For clinical management, early biomarkers could be predictive of later damage, enabling alteration of the dose to the organs at risk, use of medical countermeasures, or implementation of an appropriate long-term medical management strategy. For a nuclear exposure incident, the dose will have been delivered such that the injury falls within delayed effects of acute radiation exposure (DEARE), but the mitigator and medical-management approach would still be relevant.

In addition to identifying blood-based signatures for rapid triaging, we are also working on identifying expression changes within organs (e.g., heart, lungs, liver) affected by radiation to predict both short- and long-term organ injury. It is therefore clinically relevant to determine the response of the markers to fractionated radiation and in the presence of pre-existing conditions. Palayoor et al. showed more pronounced miRNA and mRNA expression changes in an in vitro HCAEC model after multifractionated radiation compared to a single dose of radiation suggesting that we may see similar trends in vivo [42]. Additionally, a study of acute lymphocytic leukemia patients that received six fractions of 2 Gy TBI showed a significant and sustained increase in blood-based Cdkn1a after each fraction [127]. Blood from these pediatric cancer patients also showed higher baseline levels of Cdkn1a expression compared to healthy controls, demonstrating the importance of addressing the effects of confounding factors on expression changes. Recognizing that heart biopsies would not be a suitable method to triage patients in an exposure scenario, we are also currently investigating the short- and long-term circulating RNA response of non-human primates exposed to whole thorax irradiation. This would not only enable monitoring of organ-specific damage sustained during accidental exposures but would also have applications in predicting normal tissue toxicity as a side-effect of RT for the treatment of cancer.