Changes in miRNA in the lung and whole blood after whole thorax irradiation in rats

We used a rat model of whole thorax x-ray irradiation to profile the microRNA (miRNA) in lung and blood up to 4 weeks after radiation. MiRNA from normal and irradiated Wistar rat lungs and whole blood were analyzed by next-generation sequencing and the changes by radiation were identified by differential deRNA-seq 1, 2, 3 and 4 weeks after irradiation. The average total reads/library was 2,703,137 with a mean of 88% mapping to the rat genome. Detailed profiles of 100 of the most abundant miRNA in rat blood and lung are described. We identified upregulation of 4 miRNA, miR-144-5p, miR-144-3p, miR-142-5p and miR-19a-3p in rat blood 2 weeks after radiation that have not previously been shown to be altered after radiation to the lung. Ingenuity Pathway Analysis identified signaling of inflammatory response pathways. These findings will support development of early detection methods, as well as mechanism(s) of injury and mitigation in patients after radiotherapy or radiological accidents.

(387 miRNAs) at 3, 12 and 26 weeks. However, miRNA profiles from blood or tissues that can be obtained by minimally invasive methods (such as body fluids) were not investigated in this study 14 .
Therefore the aims of the current project were: (1) to provide comprehensive profiles of the lung and blood miRNA in rats by next-generation sequencing to detect molecular changes by radiation; (2) to look for circulating miRNAs in whole blood to develop candidate biomarkers for radiation injuries. MiRNA changes were analyzed weekly up to 4 weeks after radiation, which is one week before we can intervene to mitigate pneumonitis with an ACE inhibitor 9 , and months before pulmonary fibrosis develops.

Results
Total profile of miRNA from lung and blood. Total RNA was used to generate small RNA libraries (n = 36) by size selecting for miRNA (Table 1 and Methods). Each library contained miRNA pooled from 3 rats and a total of 3 libraries were analyzed for each time point 1, 2, 3 and 4 weeks after irradiation. Separate libraries were prepared from blood and lung, each represented by 9 rats/time point. The average total reads/library was 2,703,137. The average total unique reads/library was 54,268. The percent reads that mapped to the rat genome 15 ranged from 64.7 to 98.9%, with a mean of 88%. Detailed data are shown in Table 1. A total of 947 miRNAs were identified by the next generation sequencing and 458 of them were mature unique miRNAs that have been identified in rats while 116 were homologous unique miRNAs that have been identified in other species. Another 373 miRNA were considered as new miRNA (Supplemental Table 1) that have not been previously described. We chose to follow abundantly expressed miRNAs that had more than ~100 reads in blood and ~1000 reads in lung of the control rats (Table 2). With these thresholds, we identified 113 most expressed miRNAs in blood and lung of control rats (0 Gy) ( Table 2). Among these miRNA, 43 of them from blood and 47 of them from lung were identified as new miRNAs (Table 2). In both lung and blood, 57 of the top 113 detected miRNAs were mature unique miRNAs (Fig. 1). Top 20 miRNA from lung and blood are shown in Table 3 along with results from other studies. The results from bovine blood 16 were derived by a similar protocol as the rat blood, miRNA-seq. The rat lung miRNA in the previous study 17 was obtained from a microarray and not by sequencing and so could be expected to show differing results. Since there was a large fold difference in expression of rno-miR-486-5p and 451-5p in rat blood but not bovine blood, we carried out RT-qPCR for a sub-set of the top 10 miRNA in whole rat blood without amplification or any additional experimental intervention. Amplification has the potential to alter ratios of expressed miRNA during library preparation. Expression of miR-486-5p by RT-qPCR was in the same range as miR-451-5p suggesting reads for miR-486-5p were detected at a much higher ratio by miRNA-seq in rat blood. The expression of miR-486-5p by RT-qPCR was found to be 9.5 fold higher than that of miR-191-5p in rats. Though miR-16-5p was not among the top 10 miRNA in bovine blood 16 but is abundant in human erythrocytes 17 , we checked expression by RT-qPCR and found it was also in the same range as miR-451-5p and 486-5p. Changes in miRNA profile by radiation. The numbers of the miRNA changed after radiation in blood and lung at different time points from 1 week to 4 weeks are shown in Table 4. The total reads we identified in the lung were more than those in the blood. At 1, 2 and 3 weeks the lung was found to have more miRNAs that were changed by radiation than blood (62 vs. 2; 24 vs. 13; 22 vs. 5 increase at 1, 2 and 3 weeks respectively; 40 vs. 3; 16 vs. 3; 64 vs. 15 decrease at 1, 2 and 3 weeks respectively). However, numbers of miRNA changed at 4 weeks after radiation were similar in lung and blood (4 vs. 5 increase and 8 vs. 10 decrease). Some miRNA remained changed at more than one time-point after radiation. The numbers of the changes with overlaps at different time points of the same miRNA are also listed in the Table 4. The most overlap of the upregulated miRNA in lung was at 1 and 2 weeks, with an overlap of 11 miRNAs, while 10 miRNAs at 2 weeks and 11 at 3 weeks were also decreased at 1 week.
Identification of circulating miRNA altered by radiation. Since the most miRNA changes after radiation in whole blood were shown to be at 2 weeks (13 increased) and 3 weeks (15 decreased) (Table 4), we focused on the 2-week time point to develop circulating miRNA markers. MiRNAs with statistical significance based on read counts from the blood libraries (p < 0.05 between 15 Gy at 2 weeks and 0 Gy at 1 week on the reads) were further tested by RT-qPCR (Fig. 2a). We focused on miRNA that increased after radiation since a decrease in miRNA may reflect the fall in circulating blood cells at 2 weeks. We verified the up regulation of these 4 miRNAs, which are miR-142-5p, 144-3p, 144-5p and 19a-3p by radiation (Fig. 2a). We also tested miRNAs by RT-qPCR that were reported by others to change after radiation 13 and confirmed miR-150-5p was down regulated and miR-21-5p 18 was up regulated by radiation (Fig. 2b). Sequencing results of these two miRNAs in two week blood samples followed the same trend as the RT-qPCR, though the results did not reach statistical significance between the read counts from non-irradiated controls (0 Gy) versus 15 Gy. The read counts are given in Fig. 2c. All the miRNA that we verified to be changed by RT-qPCR at 2 weeks after radiation in blood were further tested in the lung tissue at 2 weeks. Except 150-5p which showed the same change (decreased after radiation), none of them were changed in lung (data not shown). Other miRNAs described in the literature, but which did not show changes in our rat model in blood between controls and radiated rats at 2 weeks were 99b-5p, 30a-5p, let7-5p 9a-5p and 92b-3p (data not shown). The results of the RT-qPCR on these miRNAs matched our findings from sequencing.
Changes in circulating blood cells after whole thorax irradiation. The WBC are lower after one week while RBC fall by 2 weeks after radiation (Fig. 3). The WBC recover by 4 weeks.
Pathway analysis of miRNA changes in blood. The changes observed in the 2 week sequencing data from blood were also used to explore signaling pathways that are altered by radiation using the Ingenuity Pathway Analysis platform. The results showed that the top diseases and functional pathways that can be associated with the blood at 2 weeks after radiation were inflammatory disease, inflammatory response and connective tissue disorders (Fig. 4).

Discussion
In this study, we report the miRNA profile obtained by RNA-seq from rat lung and blood at baseline and 1, 2, 3 and 4 weeks after 15 Gy whole thorax irradiation. This dose induces lethal pneumonitis in rats after 42 days, though all rats survive 12 Gy whole thorax irradiation 19 . Doses of 13 or 14 Gy induce intermediate levels of lethality 19 . We used RT-qPCR to verify changes in levels of at least 6 circulating miRNA that were identified by RNA-seq at 2 weeks after 15 Gy. At this time point all rats are healthy with normal breathing rates and lung histology though we have measured an increase in vascular permeability and apoptosis in the radiated lungs 20 . Some of the miRNA that were altered in the circulation e.g. miRs 142-5p, 150-5p and 21-5p have been described in anti-apoptotic responses and may signify cellular compensation for increased hematopoietic death induced by radiation [21][22][23][24] .
The miRNA molecules we sequenced were sized by gel purification to exclude RNA other than miRNA. Our results show miRNA-seq to be a sensitive technique, yielding a dynamic range of read counts from 1 − 1.9 × 10 6 (Supplemental Table 1). Additionally 88 and 98% mean reads mapped to the rat genome in the lung and blood samples respectively ( week after radiation in lung tissue. MiRNA changes at earlier time points have been evaluated in other models of radiation but not with exhaustive techniques such as miRNA-seq. High-throughput technology such as miRNA-seq has many advantages including generation of large data sets which reflect the biology of the system. However, our study, in which expression of miR-486-5p may be overrepresented, also demonstrates limitations. Though most of the top 10 miRNA species which we identified as expressed in whole blood corresponds with another report 16 Table 2. The 113 most highly expressed microRNAs in blood and lung of control rats (0 Gy) were listed. These microRNAs showed more than 100 and ~1000 reads in blood and lung of the control rats respectively. Read count and percentage of the reads of individual microRNA in the sum of top 113 are also provided. confirmed by independent techniques/platforms regardless of specific technology or methods. We validated candidate circulating miRNAs from the high throughput RNA-seq individually by RT-qPCR. Nine miRNA showing increase from control after radiation were selected for verification by RT-qPCR. Though most miRNA trended in the same direction in both assays only 4 miRNA were statistically increased after RT-qPCR. The profile of miRNA in normal rat blood and lung (0 Gy, without radiation) that we report in this paper (Tables 2 and 3, Fig. 1) and elaborated in the supplement, shows new species that fit the definition of miRNA but remain to be characterized. More than 100 miRNA from blood and 1000 from lungs of non-irradiated rats yielded read counts greater than 100 (a value that can be consistently and easily verified by RT-qPCR). Both blood and lung tissue had 57 mature miRNAs that have been previously identified in the top 113 detected (Fig. 1).   16 , read counts were presented in reads per million (rpm) and n = 9; in the study of Caruso et al. 17 , counts were presented as absolute levels and n = 5; in the current study, total read counts were shown and n is 6 libraries, each of which included pooled RNA from 3 rats.  Table 4. Number of microRNA changed after radiation. ↑ : increase after radiation; ↓ : decrease after radiation. w: week. p < 0.05 radiated vs. non-irradiated control. Egs: 62 microRNAs were found to be increased 1week after radiation in lung; 11 of the 62 were also increased at 2 weeks after radiation in lung, 5 of the 62 were also increased 3 weeks after radiation in lung. 24 microRNAs were found to be increased in lung at 2 weeks after radiation, of them 11 was also increased at 1week, as mentioned above, and 1 microRNA was also increased at 3 weeks, and 1 also increased at 4 weeks.
Scientific RepoRts | 7:44132 | DOI: 10.1038/srep44132 Consistent with the previous findings 25 , high expression of miR-486-5p in red blood cells was also detected in whole rat blood. This miRNA was described to regulate normal erythropoiesis 26 and was also suggested to be an important oncogene in several hematopoietic myeloid malignancies 26 . MiR-486-5p was reported to be a potential plasma-based biomarker for lung cancer 27 . Leuenberge et al. showed that its abundance remained unchanged after blood transfusion in healthy males 28 . Therefore, it was used as an endogenous control for data normalization in blood miRNA studies 28,29 . In the current study, we observed decrease in miR-486-5p read-counts in rat blood MiR-191-5p was used as reference because it shows consistent read counts in all sequenced libraries. miRNA in (a) were selected by results of miRNA-seq for candidates with p < 0.05 radiated vs. control by t test. miRNA in (b) were selected because other investigators previously identified these to be altered after radiation or lung injury. Their reads count with p values from the sequence results of this study are shown in (c). # refers to the library number # in Table 1.
Scientific RepoRts | 7:44132 | DOI: 10.1038/srep44132 at 2 weeks after radiation. However these changes could not be confirmed by RT-qPCR (Fig. 2). This, along with verification of only 4/9 candidate miRNA by RT-qPCR reiterates the limitation of miRNA-seq. Using next generation sequencing, we report changing miRNA patterns after radiation in rat lung and blood ( Table 4). All RNA used for these comparisons were prepared at the same time with identical reagents and equipment. Such data can provide useful algorithms to develop early biomarkers for treatment plans in radiotherapy. As an adjunct diagnostic method, next generation sequencing provides abundant information. We verified an increase of miR-142-5p, 144-3p, 144-5p and 19-3p in blood after 15 Gy in our rat model. Another two interesting miRNA (miR-150-5p and 21-5p) were selected for testing by RT-qPCR based on changes after radiation as reported by others. MiR-150-5p was shown to be decreased after radiation in a mouse model 13 . It is also expressed in blood and hematopoietic cells, which, as mentioned, are sensitive to radiation, explaining the fall in circulating levels of this miRNA after 15 Gy whole thorax irradiation. MiR-21-5p is a well-known lung injury marker [30][31][32] . But studies with changes in miR-21-5p after radiation have been limited to lung cancer cells 23,33 and not in irradiated lung tissue. However some non-radiation studies have associated changes in this microRNA. Circulating miR-21 was suggested by Loboda et al. to be a potential early biomarker of renal fibrosis, because it was shown that upregulation of miR-21 alters metabolic pathways and leads to renal fibrosis 34 . MiR-21 was also suggested to plays a dynamic role in inflammatory responses 34 . Since inflammation and fibrosis are considered as major radiation responses in many organs including lungs, it is possible that miR-21 also plays some roles in radiation response in normal tissues. Beside circulation miR-21, exosomal miR-21 was also suggested to have a strong potential to be used as a universal biomarker to identify cancers 35 . In fact, based on one report, exosomal miR-21 seemed to be better than circulating miR-21 to serve as such a biomarker 35 . We confirmed a decrease in 150-5p and increase in 21-5p after radiation by RT-qPCR in whole blood from our rat model of thorax radiation. Though these miRNA trended in the same direction with RNA-seq data, there was considerable variation between samples in the libraries, so that the results were not statistically significant. As already mentioned, since the libraries were amplified and pooled these variations could be artifacts. We focused on the changes of miRNA at 2 weeks, since not only as mentioned, that was the time when the most increases in miRNA were detected, but also when most of the mature miRNA were found to be changed after radiation. In addition, we have also established a SPECT/CT biomarker to predict radiation lung injuries 36 . By this method, the most significant lung injuries, namely the pulmonary cell death measured by 99mTc-labeled Duramycin and pulmonary vascular resistance and vascular permeability measured in isolated perfused lungs were found to be at 2 weeks after 15 Gy in the same rat model.
MiR-144 has been reported to affect sensitivity in radiotherapy by promoting proliferation, migration and invasion of breast cancer cells 37 . Dysregulation of miR-144 has been described in studies of multiple cancers [38][39][40][41] . It was suggested to be a tumor suppressor by Matsushita et al. 38 and inhibited proliferation but promoted apoptosis and autophagy through targeting the p53-induced glycolysis and apoptosis regulator TIGAR 40 . Besides the current study, elevation of miR-144 was also reported in many non-cancer studies 42,43 . MiR-144 is overexpressed in peripheral blood mononuclear cells from patients with pulmonary tuberculosis (TB) and regulated anti-TB immune response in T cells 42 . Hassan et al. found increase in miR-144 in human bronchial epithelial (HBE) cells exposed to cigarette smoke extract and cadmium. Su et al. showed that miR-144 regulates hematopoiesis and vascular development by repressing expression of meis1 in zebrafish 44 . In summary miR-144 plays a role in inflammation, immune response and suppression of cell growth, processes that are activated by radiation. It is surprising that we observed increase in this miRNA in whole blood, since circulating cells are decreased by radiation and this miRNA was not increased in irradiated lungs in our model.
MiR-19a-3p has been reported to inhibit breast cancer progression and metastasis by inducing macrophage polarization 45 . It was also suggested to be involved in inflammatory processes by directly regulating 5-LO (5-lipoxygenase) expression in T lymphocytes 46 . Busch et al. found that the inhibition of miR-19a-3p with an antagomir led to an increase in 5-LO mRNA expression in T lymphocytes 46 . We speculate that this miRNA may be induced in response to inflammatory signaling after radiation to down-regulate lipoxygenase metabolites in blood cells such as macrophages and T-lymphocytes.
MiR-142-5p has also been demonstrated to be involved in inflammation 21,22,47 . Su et al. showed miR-142-5p regulates human and mouse macrophage profibrogenic gene expression in chronic inflammation and models of liver and lung fibrosis 47 . Increases of miR-142-5p were found in lungs of patients with idiopathic pulmonary fibrosis 47 . MiR-142-5p was reported to control T-cell responses in vitro and in murine models of graft versus host disease 22 . Finally miR-142-5p had tumor-suppressive effect in lung cancer cells 48,49 . In summary, similar to miR-144, miR-142-5p regulates cell growth and inflammation. Interestingly miR-142-5p expression in macrophages also influences fibrosis in the lung 47 , a phenotype that is induced by radiation. It is possible that detection of increase in this miRNA in spite of a decrease in circulating cells in the blood at 2 weeks after radiation suggests regulation of genes that could lead to the later effects of radiation. Perhaps these immune cells may be responsible for radiation pulmonary injury after infiltration into the lungs.
We also conducted pathway analysis to determine signaling changes in the lung after radiation, based on the changes in miRNA as determined by miRNA-seq. The pathways with the highest scores by Ingenuity Pathway Analysis were "cancer, organismal injury" and "abnormalities and reproductive system disease", while inflammatory responses received a lower score (results not shown). Pathways derived from changes that we confirmed in the blood of irradiated rats are shown in Fig. 4. To our knowledge this is the first time that miR-144-3p, 144 -5p, The pathway map is presented below and miRNA that were significantly changed in the miRNA-seq study and also validated by qPCR are shown as highlighted in red. and 19a-3p are upregulated in rat blood 2 week after irradiation of the thorax. Their function in inflammatory responses is suggested by the Ingenuity Pathway Analysis (Fig. 4). These miRNA were not changed in the irradiated lung, strongly suggesting they may be upregulated in other cells or tissues. We cannot rule out their presence in exosomes from endothelial cells within the lung. They could also be present in exosomes from other irradiated organs, e.g. the heart, partial bone marrow and blood within the thorax, that were in the field of radiation. We know immune cells are involved in radiation-pneumonitis with inflammatory infiltrates detected in the lung by histology and other diagnostic methods 50,51 . Vascular damage and remodeling are also found during radiation pneumonitis in human and animal lungs 51,52 .
Decreases in numbers of circulating blood cells are anticipated and were measured after whole thorax irradiation due to the volume of bone marrow in the field of exposure. WBC have shorter half-lives than RBC so their numbers fall before the RBC. The numbers of WBC recover by 4 weeks. We have observed increases in circulating miRNA we report after radiation, which could also be due to increase in specific transcription induced by radiation in blood cells, but not from changes in numbers of circulating cells, since these numbers were lower at the time points we examined. Future investigation with models involving radiation to the lower hemi-body or heart alone will help to highlight the effect of radiation to the lung on these miRNA changes.

Methods
Animal care and irradiation. All animal protocols and euthanasia criteria were approved by the Institutional Animal Care and Use Committee (IACUC). All methods were performed in accordance to the IACUC guidelines and regulations. Radiation was performed as described previously 53 . In brief, un-anesthetized 9-to 10-week-old female WAG/RijCmcr rats weighing approximately 140 g were immobilized in a plastic jig and irradiated with 320-kVp orthovoltage system X-rays, with a half-value layer (HVL) of 1.4 mm Cu. Rats were treated with a single dose of 15 Gy to the whole thorax at dose rate of 1.43 Gy/min. The radiation dose was delivered by two equally-weighted lateral beams to improve uniformity. The whole lung, heart and a small amount of liver were in the field. One group of age-matched rats was not irradiated (non-irradiated controls or 0 Gy group) but maintained under identical conditions. Experiments were terminated at 1, 2, 3 or 4 weeks. Age-matched controls were included at the 1 and 4 week time-points. RNA isolation. Lung tissue from the right superior lobe was cut and ~100 mg was weighed and immersed in TRIzol Reagent (Ambion/RNA by Life Technologies) immediately after sacrifice. Whole blood (0.5 ml) collected by cardiac puncture in an EDTA coated needle and syringe was directly used to extract RNA immediately without any further processing. TRIzol Reagent (Ambion/RNA (Life Technologies)) was used. Total RNA from lung and whole blood were isolated following the protocol provided by the manufacturer. RNA was stored at − 80°C for future use.
Small RNA library preparation and sequencing. Total extracted RNA was quantified using Nanodrop 2000 (Thermo Scientific). A 260/280 was used to ensure the RNA quality. Equal amounts of RNA from three rats were pooled. The range of A 260/280 was maintained between 1.6 and 2.1. Samples with ratios below 1.6 were discarded. Small RNA libraries (36 totals) were generated with the TruSeq Small RNA Library Prep kit (Illumina) following the manufacture's instruction with minor changes (Table 1). In brief, 1 μ g of the pooled total RNA was ligated with 3′ and 5′ adapters. Reverse transcription followed by PCR (based on the 3′ and 5′ adapter sequences for 16 cycles) was used to create cDNA. The amplified cDNA was purified using 6% PAGE Gel and bands between 147nt and 157nt which contains RNA fragment of 22nt and 30nt corresponding to miRNA were cut out using size markers and concentrated by ethanol precipitation. Libraries were visualized and quantitated with Agilent Technologies 2100 Bioanalyzer using DNA-1000 Chip. Next generation sequencing was performed at the Human and Molecular Genetics Center Sequencing Core at MCW 15 . In brief, prepared libraries were loaded onto the flow cell using a cBot instrument (Illumina) and clusters were generated in the flow cell after a 4-hour-PCR amplification. The flow cell was loaded onto the HiSeq 2000 sequencer. The obtained reads were fed into the in-house CASAVA informatics pipeline, where they were de-multiplexed and aligned to a reference genome miRbase v19.
Analysis of small RNA deep sequencing data. Before the analysis, the adapter sequences were first removed from the output sequence reads by the tool "trim galore" 54 . Sequences with low quality (base quality < 13) at both ends of reads were further trimmed by the FastQC 55 and mapped against miRBase v19 to identify known miRNAs using Bowtie 56 , including rat miRNAs and homologs of miRNAs known in species other than the rat. Sequence reads that did not map to miRBase were then mapped against mRNA database, Rfam (for other noncoding RNA), and RepBase (for repetitive elements) to remove reads corresponding to transcribed sequences that were not miRNAs. The remaining reads were used to predict new miRNAs with miRanalyzer 57 . miRanalyzer employs a machine learning approach based on the random forest method. With the default parameter setting, miRanalyzer can obtain the area under the curve value of 97.9% with a true positive rate of 0.79 and a false positive rate of 0.007 for predicting new mammalian miRNAs. To normalize and test differential expression, we used number of reads of known and newly identified miRNAs as input for the Bioconductor DESeq package 58 . DESeq uses a negative binomial distribution to model reads of miRNAs and to test for differential expression in deep sequencing datasets. The Benjamini-Hochberg method was used to control false discovery rate (FDR) in all statistical tests 59 .
Real-time quantitative RT-qPCR. Validation of changes in miRNA expression of selected markers was performed by RT-qPCR from the blood of rats used for miRNA-seq as well as from independently irradiated and control rats. Total RNA from each rat were required to meet the following quality criteria: 1) A 260/280 > 1.6, 2) yield of RNA was enough to carry out all the RT-qPCR reactions needed for all the primers tested. There was no pre-amplification step prior to PCR. LNA-primers were obtained from Exiqon and reactions were carried out after reverse transcription using miRCURY LNA TM Universal RT microRNA PCR kit from Exiqon with sybr green mastermix from Biotool. The relative expression of each miRNA to a reference miR-191-5p was calculated by the formula: 2 −∆CT (∆ C T= C T (target-reference). Data was then normalized to the mean relative expression of controls, to determine the fold change after radiation. MiR-191-5p was used as reference for normalization because it shows consistent read counts in the libraries by sequencing.
Blood counts. Blood was collected by cardiac puncture into EDTA tubes. White blood cell and red blood cell counts were performed by the Marshfield Laboratories the same day after checking for adequate volume and absence of clots (Marshfield, WI). Statistical analysis. Sigmaplot software was used to perform statistical analysis on the RT-qPCR results and the blood cell count results. For RT-qPCR, t test was used on 19a-3p and 21-5p; Mann-Whitney U test was used on 142-5p, 144-3p, 144-5p and 150-5p, when the data failed either normality test or equal variance test. For blood cell count, Kruskal-Wallis One Way Analysis of Variance on Ranks Multiple Comparisons versus Control Group (0 Gy) was conducted. Dunn's method was used as post hoc test.