An ionising radiation-induced specic transcriptional signature of inammation-associated genes in human whole blood: a pilot study

This communication reports the identication of a new panel of transcriptional changes in inammation-associated genes observed in response to ionising radiation received by radiotherapy patients. Peripheral blood samples were taken with ethical approval and informed consent from a total of 20 patients undergoing external beam radiotherapy for breast, lung, gastrointestinal or genitourinary tumours. Nanostring nCounter analysis of transcriptional changes was carried out in samples prior and 24 hours post-delivery of the 1st radiotherapy fraction, just prior to the 5th or 6th fraction, and just before the last fraction. Statistical analysis with BRB Array Tools, GLM MANOVA and nSolver, revealed a radiation responsive panel of genes which varied by patient group (type of cancer) and with time since exposure (as an analogue for dose received), which may be useful as a biomarker of radiation response. Further validation in a wider group of patients is ongoing, together with work towards a full understanding of patient specic responses in support of personalised approaches to radiation medicine.


Introduction
A variety of different biological and physical retrospective tools are available to assess individual radiation doses following a radiation accident or incident (1)(2)(3)(4). In recent years, transcriptional changes in blood have been identi ed as a promising biomarker for radiation response to support biodosimetric assessment of individual doses in accidental exposure scenarios (5,6). The responsiveness of FDXR to ionising radiation at the transcriptional level in human blood was recently reported to provide accurate in vivo dose estimates and providing the rst in vivo dose response in humans (7). Transcript variants of this gene have also shown a remarkable potential as standalone biomarkers for ionizing radiation exposure screenings (8). The in uence of several potential confounding factors (cancer condition, sex, simulated bacterial infection (lipopolysaccharide), and curcumin, an anti-antioxidant agent) on radiation dose estimation using in vivo validated transcriptional biomarkers was also investigated, with the outcome that such confounding factors should not prevent the use of transcriptional responses for emergency triage purposes (9). Most recently, a new protocol for rapid gene expression-based dose estimation in human blood was reported (10), together with the generation of a transcriptional radiation exposure signature in human blood using long-read nanopore sequencing including several new genes representing ideal biomarkers of radiation exposure (11).
Although radiation-responsive genes such as FDXR have been reported in human white blood cells in radiotherapy patients, they have limited inter-individual variability in response; apart from early responsive genes such as CDKN1A which may be help to predict the severity of acute skin radiation toxicity (12), they are not informative of inter-individual variability in normal tissue sensitivity to radiation exposure. We previously reported that several genes associated with in ammatory processes (ARG1, BCL2L1, and MYC) present a long-term modi cation of transcriptional expression (13) towards the end of the radiotherapy treatment. In addition, there is strong evidence of radiation-induced in ammation feeding into innate and adaptive antigen-speci c immune responses. This adds another dimension to the tumour-host crosstalk during radiation, in turn in uencing normal tissue side effects following radiotherapy (14)(15)(16)(17)(18).
In this communication, we present a panel of in ammation-associated genes which are radiation responsive, validated in a population of radiotherapy patients, and shown to be differentially responsive for different patient groups and at different time points post-exposure.

Methods
The general conditions for ethical approval, patient selection and informed consent, blood sampling and individual patient dosimetry were described in detail in Moquet et al., (2018) (19). Brie y, blood samples from ve breast, four endometrial, ve lung, three prostate, two oesophagus and one colon cancer patients, treated with Intensity Modulated Radiotherapy (IMRT) using a linear accelerator (LINAC) were collected at four different time points during the course of the treatment: before the start of the treatment, 24 h after the rst fraction, just before the fth or sixth fraction and the last fraction. The prescribed doses for each patient were as follow: breast cancer patients received 40-40.5

RNA isolation and reverse transcription
Total RNA was extracted with the PAXgene Blood miRNA kit (Qiagen, PreAnalytiX GmbH, Hilden, Germany) using a robotic workstation Qiacube (Qiagen, Manchester, UK). The quantity of isolated RNA was determined by spectrophotometry with a ND-1000 NanoDrop and quality was assessed using a Tapestation 220 (Agilent Technologies, CA, USA). cDNA was prepared from 350 ng of the total RNA using High Capacity cDNA reverse transcription kit (Applied Biosystems, FosterCity, CA, USA) according to the manufacturer's protocol.
nCounter analysis Samples were analysed by the nCounter Analysis System (NanoString Technologies®, Inc., Seattle, WA, USA) according to the manufacturers' guidelines. The samples were run using 100 ng RNA per sample on the Human In ammation V2 panel, which consists of 249 genes.

BRBArrayTools and MANOVA
Statistical analysis was performed with BRBArrayTools, to identify genes for which there were statistically signi cant changes (up or down regulation; p < 0.05) associated with number of radiotherapy (RT) fractions and time since exposure, with a false discovery rate (FDR) < 0.05. General Linear Model Multivariate Analysis of Variance (GLM ANOVA) and Multivariate Analysis of Variance (MANOVA) was then carried out with Minitab18®, to identify panels of genes signi cantly associated (p < 0.05) with radiation exposure, taking into account radiotherapy patient group by type of cancer treated (breast, lung, gastrointestinal or genitourinary tumours) and time since exposure (just before exposure, 24 hours postexposure, and just before the nal fraction -range 3-5 weeks for all patients).

Differential expression and pathway analysis
Nanostring nCounter nSolver 4.0 (Nanostring Technologies) with the Advanced analysis plugging (version 2.0.134) was used to perform the differential expression (DE) and pathway analysis. DE analysis includes several multivariate linear regression models to identify signi cant genes (mixture negative binomial, simpli ed negative binomial, or log-linear model). FDR p-value adjustment was performed with Benjamini-Yekutieli method (21). Statistically signi cant, differentially expressed genes were de ned as those with expression levels corresponding to a log2 ratio >0.5 or < −0.5 and p-value < 0.05.
Gene set analysis (GSA) is a quantitative summary of DE for gene sets. Gene set's global signi cance score for a covariate is calculated as the square root of the mean squared t-statistic of genes. Global and directed signi cance scores were calculated for each pathway. Pathway scores were used to summarize data from a pathway's genes into a single score. Pathway scores were calculated as the rst principal component of the pathway genes' normalized expression and standardized by Z scaling. Pathway scoring helps to see how pathway scores change across samples. Increasing score corresponds to mostly increasing expression.
Immune cell type pro ling The cell type pro ling module in Nanostring nCounter nSolver 4.0 advanced analysis was used to quantify cell populations using marker genes. Raw cell type measurements are calculated as the log2 expression of each cell type's marker genes and show the estimated abundances of each individual cell type between samples.

BRBArrayTools and MANOVA
A total of 29 genes were identi ed by BRB-ArrayTools as being signi cantly down or up regulated in response to ionising radiation exposure, with FDR < 0.05. GLM ANOVA then revealed a subset of downregulated genes only for which both time since exposure and type of cancer were statistically signi cant, which indicates that these genes may be informative in understanding patient group speci c responses.
MANOVA on these genes reveals that this combined set of 7 genes (MYC, CD40LG, CCL4, IL7, TCF4, CCR7 and FASLG) is together statistically signi cantly reliant on both time post-exposure (p = 0.042) and cancer type (p < 0.001). For the up-regulated genes, no signi cant effects were identi ed for time postexposure.
Differential Gene expression analysis DE analyses revealed a set of genes differentially expressed in blood samples at the time point before the last fraction ( Figure 1). The genes TLR8, ALOX5, TYROBP, MAPK1, MYD8B, BCL6, HHGNI, MYC and MAPKAPK5 presented an up-or down-regulation (p-vale<0.05) independently of the cancer type ( Figure   1C). However, only MYC and BCL6 showed a fold change regulation above 1.5-fold (log2 FC >0.5 or < −0.5).

Gene set analysis (GSA)
Differentially expressed gene sets were observed only when comparing the before time point with the last time point (before last fraction) (Figure 2A). These results highlighted the Interleukin 18 family and Class I MHC mediated antigen processing & presentation genes, apoptosis, interleukin 20 family, platelet homeostasis and defensins pathways with the highest scores (Figure 2 A). Directed global signi cance scores indicated that these pathways are upregulated at the last time point (Figure 2 B).

Pathway scoring
Pathway score clustering showed a general separation of the time points for some of the cancer types, indicating higher scores at the last time point and also for breast cancer patients (Figure 3.A). When we compared pathway scores to time points ( Figure 3B) we observed a decrease in NF-kB, cell cycle and apoptosis at the last time point. However, the scores in interleukin 1 signalling, Class I MHC mediated antigen processing & presentation genes, cellular senescence, signalling by FGFR4, C-type lectin receptors and MAPK1MAPK3 signalling pathways were increased in the last time point ( Figure 3B).
Regarding cancer type, interferon signalling pathway had a high score in breast cancer and low in colon and oesophagus compared to the other cancer types. Low scores were also observed for interleukin 1 signalling and metabolism pathways in colon. In contrast, the defensins pathway presented a high score in colon and oesophageal cancers.
Pathway scoring between only the rst and last time points showed two clusters one of higher scores in the last time points for a group of samples and a low score cluster for the rst time point (before time point) which shows a switch on of several in ammation related pathways at the end of the radiotherapy treatment.

Immune Cell pro ling
Cell type pro ling analysis identify two main cell populations, macrophages and exhausted CD8+ T cells (T cells which adopt a functionally attenuated state due to prolonged antigen stimulation, characteristic of chronic infections and cancer). The average cell type score was compared between the different time points and rst and last time points ( Figure 4A). The results showed that macrophages were found to be relatively higher before the start of the treatment compared to the last time point but the opposite was found for the exhausted CD8 T cells ( Figure 4A).
When we compared the cell type scores between the different cancer type ( Figure 4B), the results revealed that breast and oesophagus cancer patients presented opposite changes in these two cell types compared to the other cancer types. Both cell types were higher in breast cancer patients but lower in oesophagus cancer patients compared to colon, prostate and lung cancers.
Validation of nCounter analysis by qPCR IL7 and CD40LG were selected to validate the nCounter analysis ( Figure 5). The expression pro les of these genes were con rmed by qPCR with a signi cant down-regulation at the last time point (last fraction of the radiotherapy treatment).

Discussion And Conclusions
Our group has previously reported long-term modi cation of transcriptional expression in genes associated with in ammatory processes in head and neck and endometrial cancer patients undergoing radiotherapy (13).The aim of this study was to report, for the rst time in a wide range of cancer types (breast, lung, prostate, endometrium and gastro-intestinal), details of a panel of in ammation-associated genes identi ed in radiotherapy patients as being signi cantly associated with ionising radiation exposure. Radiation responsive genes, both up and down regulated, were identi ed using a Human In ammation V2 panel from nCounter Analysis system (NanoString Technologies®) and assessed for the signi cance of their individual and combined responses in terms of time since exposure and type of cancer. The radiation doses received varied on a patient by patient basis (full data in Moquet et al., 2018 (19)), however, for the purposes of this initial analysis, as each time point pre-and post-exposure was at a different stage of the treatment, time post-exposure can be taken as an analogue for dose. Thus, the results of this work reveal gene sets and pathways which show signi cance in terms of radiation responsiveness for different groups of patients irrespective of the type of cancer. Both statistical approaches performed in this study identi ed a common radiation responsive gene, MYC. MYC is a protooncogene involved in cell cycle, cell proliferation, apoptosis (22), regulation of innate and adaptive host tumour immune responses (23) and it has been previously described as a radiation responsive gene in different cohorts (head and neck and endometrial cancer patients) (13). Only MYC and BCL6 were shown to be upregulated in our differential gene expression analysis performed with nSolver advanced analysis software, whereas the MANOVA analysis identi ed MYC, CD40LG, CCL4, IL7, TCF4, CCR7 and FASLG as signi cantly differentially expressed. From those genes, CD40LG and IL-7 were observed to be signi cantly downregulated (before 5th -6th and last RT fraction) using nCounter analysis which was further con rmed by qPCR. However, in qPCR the signi cance of both CD40LG and IL-7 was observed only before last fraction among all the tumour types, demonstrating the sensitivity of these assays. CD40LG is transiently expressed on T cells as a result of in ammatory response and known to activate CD40 (24). CD40, a member of TNF family is known to be expressed by DC, myeloid cells and B cells and its activation leads to priming of cytotoxic T cells (25). Recently, IL-7 has been shown to be produced by radioresistant haematopoietic cells in mice (26). IL-7 regulates homeostasis of lymphocytes, survival and maintenance of T cells (27). These results suggest that CD40LG and IL-7 have potential as immunein ammatory markers to correlate dose fraction against volumes irradiated.
Pathway analysis revealed that there is a modulation of in ammation associated pathways after recurring exposure to radiation during the course of the radiotherapy treatment. It is known that IR can induce in ammation by inducing cytokines secretion and through bystander signals (28)(29)(30). GSA analysis highlighted two upregulated pathways at the last time point (before the last fraction), interleukin-8 and class I MHC mediated antigen processing and presentation. Interleukin-18 is involved in activation and differentiation of various T cell populations (31) and its increase has been linked to radiation injury (32). MHC class I peptides are antigens originated intracellular and delivered to the cell surface to be recognized by CD8 + T cells and an increase in this cell surface peptide presentation has been described after gamma irradiation exposure (33).
Looking at the individual cancer types, breast cancer patients presented a high score in the interferon pathway compared to the other type of cancers. Interferons are a family of cytokines which play an important role in initiating immune responses, especially antiviral and antitumour effects (34,35). For colon and oesophageal cancers, defensins pathway presented a high score. The modulation of stress and multiple immune parameters were also reported for head and neck (36) and prostate adenocarcinomas (37) treated by radiotherapy.
Cell markers were present in the nCounter panel for macrophages and exhausted CD8 + T cells. These markers revealed changes in the levels of these cell types during the radiotherapy treatment and between cancer types. Continuous radiation exposure seems to promote a decrease in macrophages at the end of the treatment. Macrophage irradiation has demonstrated to modulate their phenotype towards a proin ammatory state promoting cancer angiogenesis and cancer cell-invasion (38). Macrophage activation and recruitment at site of injury has been proposed as an indirect effect of IR which results from cellular damage signals to clear radiation-induced apoptotic cells (28). This macrophage recruitment is in line with the slight decrease of macrophages in the early time point of the radiation therapy (24h after rst fraction and before 5th -6th fraction). However, it is not clear why the macrophages decrease after long periods of repetitive exposures in the present study. When looking at individual cancer types, breast and oesophagus cancer patients differ in the level of macrophages compared to the other cancer types. Both oesophageal cancer patients had loco-regional lymph nodes included in the radiation eld whereas only one out of 5 breast cancer patients had large radiotherapy volumes with lymph nodes irradiated. It is di cult to draw any rm biological conclusions given the small sample size, but this will be explored further in the ongoing follow-on study.
T cell exhaustion is an attenuated state of cell-response resulting from repeated or prolonged antigenic exposure under suboptimal conditions (39). So, it is not surprising to see this cell group increased after the continuous exposure to IR at the end of the radiotherapy treatment. T cell exhaustion also differs between different tumour types (40), as seen in the different cancer groups.

Conclusion
In summary, we con rmed that changes in expression of speci c genes as well as systemic in ammatory responses showed great promise as markers for individual radiation dosimetry. Moreover, these results are encouraging and will be used as part of further research to understand individual radiation responses and explore the links between in ammatory and immune responses in the context of different dose fractionation schedules and volumes irradiated in various cancer types. Ultimately it is hoped these data will help further to develop personalized use of radiation in medicine. Availability of data and materials

Abbreviations
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests Funding