*Review* **Ionizing Radiation Protein Biomarkers in Normal Tissue and Their Correlation to Radiosensitivity: Protocol for a Systematic Review**

**Anne Dietz, Maria Gomolka, Simone Moertl and Prabal Subedi \***

Bundesamt für Strahlenschutz/Federal Office for Radiation Protection, Ingolstädter Landstraße 1, 85764 Oberschleissheim, Germany; adietz@bfs.de (A.D.); mgomolka@bfs.de (M.G.); smoertl@bfs.de (S.M.) **\*** Correspondence: psubedi@bfs.de; Tel.: +49-30183332244

**Abstract:** *Background:* Radiosensitivity is a significantly enhanced reaction of cells, tissues, organs or organisms to ionizing radiation (IR). During radiotherapy, surrounding normal tissue radiosensitivity often limits the radiation dose that can be applied to the tumour, resulting in suboptimal tumour control or adverse effects on the life quality of survivors. Predicting radiosensitivity is a component of personalized medicine, which will help medical professionals allocate radiation therapy decisions for effective tumour treatment. So far, there are no reviews of the current literature that explore the relationship between proteomic changes after IR exposure and normal tissue radiosensitivity systematically. *Objectives:* The main objective of this protocol is to specify the search and evaluation strategy for a forthcoming systematic review (SR) dealing with the effects of in vivo and in vitro IR exposure on the proteome of human normal tissue with focus on radiosensitivity. *Methods:* The SR framework has been developed following the guidelines established in the National Toxicology Program/Office of Health Assessment and Translation (NTP/OHAT) Handbook for Conducting a Literature-Based Health Assessment, which provides a standardised methodology to implement the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach to environmental health assessments. The protocol will be registered in PROSPERO, an open source protocol registration system, to guarantee transparency. *Eligibility criteria:* Only experimental studies, in vivo and in vitro, investigating effects of ionizing radiation on the proteome of human normal tissue correlated with radio sensitivity will be included. Eligible studies will include English peer reviewed articles with publication dates from 2011–2020 which are sources of primary data. *Information sources:* The search strings will be applied to the scientific literature databases PubMed and Web of Science. The reference lists of included studies will also be manually searched. *Data extraction and results:* Data will be extracted according to a pre-defined modality and compiled in a narrative report following guidelines presented as a "Synthesis without Meta-analyses" method. *Risk of bias:* The risk of bias will be assessed based on the NTP/OHAT risk of bias rating tool for human and animal studies (OHAT 2019). *Level of evidence rating*: A comprehensive assessment of the quality of evidence for both in vivo and in vitro studies will be followed, by assigning a confidence rating to the literature. This is followed by translation into a rating on the level of evidence (high, moderate, low, or inadequate) regarding the research question. Registration: PROSPERO Submission ID 220064.

**Keywords:** ionizing radiation; normal tissue; biomarker; radiotherapy; radiosensitivity; protein

#### **1. Introduction**

#### *1.1. Background and Rationale*

The International Agency for Research on Cancer (IARC) Global Cancer Observatory reports more than 18 million new cases of cancer in 2018 [1] and radiotherapy (RT) is used to treat 50–60% of cancers [2]. The delivered dose during standard RT is balanced between optimal tumor kill and avoidance of damage to surrounding tissues [3]. Depending on

**Citation:** Dietz, A.; Gomolka, M.; Moertl, S.; Subedi, P. Ionizing Radiation Protein Biomarkers in Normal Tissue and Their Correlation to Radiosensitivity: Protocol for a Systematic Review. *J. Pers. Med.* **2021**, *11*, 3. https://dx.doi.org/10.3390/jpm 11010003

Received: 10 November 2020 Accepted: 18 December 2020 Published: 22 December 2020

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/ licenses/by/4.0/).

the tumor entity, up to 20% of patients show a moderate to severe detrimental response to ionizing radiation (IR) treatment [4]. Acute effects include erythema, inflammation and mucositis depending on cancer type while late effects are typically fibrosis, atrophy, vascular damage and neurocognitive and endocrine dysfunctions, especially for brain irradiated children [5–7]. These side effects limit radiation doses that can be applied to the tumor, often leading to suboptimal tumor control or to serious impairment of the quality of life of survivors. In a small subset of patients the severe reactions can be ascribed to known radiation hypersensitivity syndromes, such as Ataxia–Telangiectasia (A-T), Fanconi anemia (FA), or Nijmegen Breakage Syndrome (NBS) [8–10]. As late as 2010, children with AT mutations have succumbed to death following RT [11]. These genetic syndromes, however, only comprise about 1% of the patients demonstrating severe side effects [12]. Therefore, most of the normal tissue reactions cannot be explained by known genetic disorders and no clear guidelines exist for medical doctors to individually adapt their therapy scheme. This has led to an increased interest in predicting personalized radiosensitivity.

Radiosensitivity is any enhanced tissue or cell reaction after a subject has been exposed to IR when compared to the majority of other "normal" responding individuals [13,14]. The reactions include inflammation, fibrosis, cardiovascular illness, cataracts, and cognitive decline [15]. Individual radiosensitivity can be applied as component of personalized medicine in RT. Personalized medicine is not about finding out novel medications but sub-dividing individuals in various subgroups that vary in their response to treatment for a specific disease [16]. It is performed to tailor the treatment to the individual need. This implies that medical professionals can target cancer patients, who are radiosensitive, with lower doses and alternative treatment schedules, for example, chemotherapy rather than radiotherapy. On the other hand, patients who are less radiosensitive could be given higher doses of IR to maximize the likelihood of treatment success [14]. Moreover, although the potential for therapy using ionizing radiation is unparalleled, there is an increasing concern for the risks posed by low-dose occupational exposure among workers in nuclear industries and healthcare [17–19].

It was already established in the early twentieth century that individuals respond differently to IR [20] and the reason behind this is still under thorough investigation. Discovering breakthroughs in individualized radiosensitivity is difficult because the effect of IR is modified by age, gender, lifestyle, genetic predisposition, and the quality and quantity of IR dose these individuals receive [21]. Several conventional reviews that summarize IR-induced changes at a molecular level have been published over the years. For example, a compilation of cytogenetic damage, epigenomic alterations, induced and germline mutations, DNA and nucleotide pool damage, and transcriptomic and translational biomarkers of radiation exposure for epidemiological studies have been reported [22]. Similarly, DNA double stand break repairs, chromosomal aberrations and radiation-induced apoptosis in ex vivo irradiated blood lymphocytes as predictors of radiosensitivity have also been described [23]. A review of various proteomics approaches to investigate cancer radiotherapy in cancer and normal cell lines and in bio fluids of in vivo irradiated individuals has also been performed [24].

This review is different to conventional reviews—it is a systematic review (SR). Unlike conventional reviews, SR provides an unbiased selection of studies that include an objective and transparent evaluation of the evidence. [25,26]. SR begins with defining the terms PECO—population, exposure, comparators, and outcome, which helps to produce a well formulated research question for the SR. Each of the terms has an inclusion and exclusion criteria, which furthermore specifies which studies will be included. Each study is then evaluated for the relationship between exposure and outcome, as well as dissected based on questions that define selection, confounding, performance, attrition or exclusion, detection, and selective reporting risk of biases [27,28].

Out of 28,279 studies (PubMed search term: ionizing radiation [Title/Abstract], retrieval date 23 September 2020) there are no systematic reviews (SRs) that investigate proteomic changes after exposure to IR. Taking into consideration that the study of Per-

not et al. [22] including protein biomarkers for IR exposure took place in 2012, we have compiled proteomic markers of radiosensitivity in normal tissues from the last 10 years (2011–2020). In this article we provide a protocol that determines our search and evaluation strategy for the actual systematic review.

Our planned review aims at presenting the status quo of IR-induced changes in protein expression in normal tissue that can be correlated to radiosensitivity, which can be used to further investigate the concept of individual radiosensitivity. This will help to personalize treatment strategies for cancer patients during radiotherapy (RT) or help to assist an individualized risk assessment process by identifying and protecting occupationally exposed persons l nuclear workers and radiologists. A future issue may be the protection of sensitive cosmonauts from harmful effects of cosmic radiation.

#### *1.2. Objectives*

The main objective of this SR is to evaluate the effects of ionizing radiation on the proteome of human normal tissue regarding radiosensitivity in experimental models (in vivo and in vitro).

Following sub-objectives will be taken into account:


#### **2. Methods**

This structure for this systematic review, as presented in the graphical abstract, was adapted according to the National Toxicology Program/Office of Health Assessment and Translation NTP/OHAT handbook [27], which provides standard operating procedures for conducting a systematic review and integrating evidence [26,29]. To rate the quality of the scientific evidence, GRADE (Grading of Recommendations Assessment, Development and Evaluation) will be used. This is a formal process that is often used in systematic reviews, which is also applied to develop recommendations in guidelines that are as evidence-based as possible [25,29]. The selection of articles, data extraction and synthesis, as well as risk of bias assessment, will be performed manually. The synthesis of data will be performed narratively without meta-analysis, as explained by Campbell et al. [30]. This SR will adhere strictly to PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [28,31], which provide an evidence-based minimum set of items that need to be reported for evaluation of randomized trials or can be used as a basis to judge other research types, e.g., evaluations of interventions. The protocol and the abstract will also be reported as described in PRISMA-P [32] and PRISMA-A [33] respectively.

The SR is registered in the International Prospective Register of Systematic Reviews on 10 November 2020 (submission ID 220064)

#### *2.1. Eligibility Criteria*

Studies that comply with elements of PECO (Population, Exposure, Comparators, and Outcome as outlined in Table 1) will be included in this SR.

#### 2.1.1. Population

The population in this SR will include both in vivo and in vitro models.

In vivo models: This model will include humans or blood, biopsies, and body fluids taken from humans.

In vitro models: This model will include non-cancer tissue culture, primary human non-cancer cell lines, or non-cancer cell lines derived from humans.

This review will exclude non-human studies. This review will also exclude tumour cell lines, tumour tissue, and biopsies. The rationale behind excluding tumour data is that we focus on radiation induced effects in normal tissue. Although some mechanisms and pathways may overlap, there are also clear differences of radiation resistance and sensitivity mechanisms in tumour compared to normal tissue. The protein markers identified here should help to predict radiation sensitivity reactions of normal tissue of cancer patients and therefore assist a personalized radiation therapy treatment. In addition, identified markers can help in risk assessment of radiation exposed individuals, such as nuclear workers, accidentally exposed individuals, and individuals living in areas of higher background ionizing radiation.

**Table 1.** Population, exposure, comparators, and outcome (PECO) Statement with inclusion and exclusion criteria.


#### 2.1.2. Exposure

The exposure will be ionizing radiation (IR). The World Health Organization defines IR as radiation with enough energy that during an interaction with an atom, it can remove tightly bound electrons from atoms, which results in the atom being charged or ionized [34]. Therefore, this study will include all sources of IR: X-Ray, cosmic rays, gamma ray, alpha and beta particles, carbon and proton therapy, and all sources of natural background ionizing radiation. This review will exclude non-ionizing radiation (infrared, near-infrared, ultraviolet, microwaves, electromagnetic radiation or radio waves)

#### 2.1.3. Comparators

Comparators in this study will be humans or in vitro models that have not been exposed to IR.

In case of studies including humans exposed to IR, material such as blood, before and after IR, taken from the same human, will be included. When tissue samples before and after irradiation are compared the localisation of the samples within the radiation field must be ensured and dose estimates should be provided.

In case of studies involving humans living in areas of higher-than-average natural background radiation, comparators will be humans that live in areas of average natural background radiation but from a similar demographic community.

Studies that do not have a comparator group will be excluded.

#### 2.1.4. Outcomes

Outcomes of interest are changes in protein expression levels correlated to radiosensitivity. We have explained before that radiosensitivity could include inflammation, fibrosis, cardiovascular illness, and cognitive decline. Therefore, studies that have included information on such parameters will be included.

In Vivo models should report overall survival and in vitro models should mention survival, apoptosis, proliferation, colony formation or metabolic assays.

#### 2.1.5. Exclusion criteria prioritisation

Studies will be excluded if they are:

• Not a primary study


#### *2.2. Search Strategy*

#### 2.2.1. Databases

The searches will be performed in NCBI PubMed [35] (https://pubmed.ncbi.nlm.nih. gov/) and ISI Web of Knowledge v.5.34 [36] (https://www.webofknowledge.com/). Any additional study might be added manually later. The references will be imported into Microsoft Excel and the duplicates removed. The search string for ISI Web of Knowledge is provided in Supplementary Information with this protocol.

#### 2.2.2. Search Strings

The search strings will be a combination of population, exposure, and outcome elements from the PECO parameters. The population of interest are human and/or normal tissues, the exposure of interest is ionizing radiation, and the outcome of interest is 'radiosensitivity and the corresponding changes in protein expression'. Restriction for language and time period will be set where studies published between 2011 and 2020 in English will be considered. Search strings that were used in Web of Science are presented in Supplementary Information and the strings will be adapted and calibrated to be used in PubMed.

#### 2.2.3. Study Selection

Studies will be subjected to a two-phase screening, which is also presented in the graphical abstract. As a Phase I screening, AD and PS will together cross-check the title, abstract, and the key words with the inclusion/exclusion criteria. Retained articles will be downloaded for a phase II full-text screening manually. Any article excluded in Phase II screening, along with the reason for exclusion, will be recorded and provided in the Supplementary Information. Any disagreements between the reviewers will be solved by consensus, involving MG or SM if necessary.

#### *2.3. Data Extraction*

Data extraction will be performed by PS and AD together and any discrepancies will be solved by consensus. Google sheets will be used to enter the data and the result will be finally reported in Microsoft Excel. The form for data extraction is provided in Supplementary Information.

#### *2.4. Body of Evidence Structure*

Evidence will be organised in outcome-related groups favouring data synthesis (proteins) and confidence rating at the health outcome level (radiation induced normal tissue radio sensitivity).

The criteria to determine the inclusion of specific studies or experiments in each outcome group will consider the evidence stream (i.e., in vivo, in vitro), health outcomes /endpoints or exposure regime (high or low dose, duration). Two outcomes, primary and secondary, have been defined. According to the literature, IR induces toxicity in normal tissue and in some cases shows a radio sensitive phenotype. This is the definition of primary outcome in this review. Secondary outcomes represent intermediary endpoints upstream of primary outcomes. Altered protein expression after exposure to IR, which may be grouped around specific signalling or functional pathways, has been defined as the secondary outcome.

#### *2.5. Internal Quality Assessment*

The included studies will be internally quality controlled using the Risk of Bias (RoB) tool developed by the Office of Health Assessment and Translation [27,37]. The RoB tool acts as an internal quality control in reviewing the articles included for the SR. Even though the studies follow a methodological flow, they might still have a bias, which might lead to an underestimation or an overestimation of the effect of the exposure. For example, if the population contains mainly older subjects and the comparator contains younger ones, the effect of the exposure might be overestimated. To critically evaluate the studies, the following questions will be asked:

#### 2.5.1. Selection Bias


#### 2.5.2. Confounding Bias


#### 2.5.5. Detection Bias


#### 2.5.7. Other Sources of Bias

11. Were there any other potential threats to internal validity (e.g., statistical methods were appropriate and researchers adhered to study protocol?)

Each question will be answered with 'definitely low risk of bias', 'probably low risk of bias', 'probably high risk of bias', or 'definitely high risk of bias'. Responses will be determined together by PS and AD. Considering the RoB responses of each question, the study will be categorized into three tiers, T1, T2 and T3, as proposed by the National Toxicology Program/Office for Health Assessment and Translation (NTP/OHAT). The tiers will be based on 'Key Questions', which are domains of randomization bias, outcome detection bias, and performance bias.

#### *2.6. Confidence Rating for Each Body of Evidence*

Extracted data from each study included will be considered as independent bodies of evidence. An assessment will be performed for each to define a confidence rating. The confidence rating reflects the reliability with which the study findings accurately depict a true effect of IR toxicity on normal tissue and linkage to radiosensitivity, as described in the NTP/OHAT handbook [27]. Each body of evidence is given an initial confidence rating that is downgraded or upgraded according to factors that decrease or increase confidence in the results.

Initial confidence rating is based on the presence or absence of four features. The features are (1) controlled exposure (2) exposure prior to outcome (3) individual outcome data and (4) use of comparison group

The ratings are as follow:


The initial confidence is then upgraded or downgraded depending on certain factors, and a confidence in the body of evidence is provided (Table 2). The factors increasing confidence are magnitude, dose response, and consistency across studies and the factors decreasing confidence are risk of bias, unexplained inconsistency, and imprecision.


#### **Table 2.** Accessing confidence in body of evidence.

#### *2.7. Translation of Confidence Rating Into Level of Evidence (for the Health Effect)*

Ionizing radiation leads to toxicity in all living organisms, in both tumour and nontumour cells. Therefore, the translation of confidence rating into levels of evidence will not be performed in the review. This SR aims to investigate proteomic changes in normal tissues and our population consists of in vitro studies of primary human material and immortalized cell lines as well as in vivo studies, on occupationally, naturally and accidentally exposed persons and on radiotherapy patients. The defined population in this SR are exposed to a broad range of radiation qualities, IR doses and dose-rates, and focusing on one particular type of dose and dose would hamper the objective of the study. Therefore, no preference for high or low-doses, respectively, or high or low-dose rates will be made. The different doses and dose-rates, however, will be provided for all selected studies in the final review.

#### *2.8. Data Synthesis*

Included data will likely comprise randomized and non-randomized trials. Moreover, clinical diversity is inevitable because of the defined PECO parameters as effects of IR on expression of different proteins being investigated. The secondary outcome is an altered protein expression after a population has been exposed to IR. It is highly probable that studies look at a diverse set of proteins, and studies will not have proteins in common that show an altered expression. Therefore, a meta-analyses might not be possible and in that case synthesis of results is performed in a narrative way or described textually. No reporting guidelines exist for narrative synthesis and although it provides a clear picture of the effects of exposure, such synthesis lacks transparency [38]. When methods other than meta-analyses are used to synthesize results, certain findings, such as reporting of synthesis structure and comparison grouping, standardised metric used for synthesis, synthesis method, presentation of data, and the summary of the synthesis finding, are left

unreported. To counter these problems, data synthesis will be reported using the narrative synthesis without meta-analyses (SWiM) method, as presented by Campbell et al. [30].

#### *2.9. Differences between Protocol and the Review*

If there are any methodological deviations from this protocol in the review to be written, they will be mentioned in the 'Differences between protocol and review'.

#### *2.10. Potential Applications of the Protocol/Review*

The application or outcome of this protocol is the definition of parameters for the systematic evaluation of protein changes which are correlated with radiosensitivity in normal tissue. The primary outcome of the systematic review is the identification of these proteins. In the review article we will also discuss potential applications of these proteins in medical practice.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2075-442 6/11/1/3/s1, data extraction form, Risk of bias questions, search algorithm for web of science and Prisma protocol checklist.

**Author Contributions:** Conceptualization, A.D., M.G., S.M., and P.S.; methodology, A.D., M.G., S.M., P.S.; formal analysis, A.D., M.G., S.M., P.S.; investigation, A.D., M.G., S.M., P.S.; resources, M.G., S.M.; writing—original draft preparation, A.D., P.S.; writing—review and editing, A.D., M.G., S.M., P.S.; supervision, M.G., S.M.; project administration, M.G., S.M.; funding acquisition, M.G., S.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Bundesministerium für Bildung und Forschung (Germany) Grant 02NUK035D (AD) and Grant 02NUK047B (PS).

**Acknowledgments:** The authors would like to thank Bernd Henschenmacher and Elisa Pasqual for the guidance in writing this protocol and Lukas Duchrow for his help with statistics.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


## *Systematic Review* **Ionizing Radiation Protein Biomarkers in Normal Tissue and Their Correlation to Radiosensitivity: A Systematic Review**

**Prabal Subedi \* , Maria Gomolka, Simone Moertl and Anne Dietz**

Bundesamt für Strahlenschutz/Federal Office for Radiation Protection, Ingolstädter Landstraße 1, 85764 Oberschleissheim, Germany; mgomolka@bfs.de (M.G.); smoertl@bfs.de (S.M.); adietz@bfs.de (A.D.) **\*** Correspondence: psubedi@bfs.de; Tel.: +49-30183332244

**Abstract: Background and objectives**: Exposure to ionizing radiation (IR) has increased immensely over the past years, owing to diagnostic and therapeutic reasons. However, certain radiosensitive individuals show toxic enhanced reaction to IR, and it is necessary to specifically protect them from unwanted exposure. Although predicting radiosensitivity is the way forward in the field of personalised medicine, there is limited information on the potential biomarkers. The aim of this systematic review is to identify evidence from a range of literature in order to present the status quo of our knowledge of IR-induced changes in protein expression in normal tissues, which can be correlated to radiosensitivity. **Methods**: Studies were searched in NCBI Pubmed and in ISI Web of Science databases and field experts were consulted for relevant studies. Primary peer-reviewed studies in English language within the time-frame of 2011 to 2020 were considered. Human non-tumour tissues and human-derived non-tumour model systems that have been exposed to IR were considered if they reported changes in protein levels, which could be correlated to radiosensitivity. At least two reviewers screened the titles, keywords, and abstracts of the studies against the eligibility criteria at the first phase and full texts of potential studies at the second phase. Similarly, at least two reviewers manually extracted the data and accessed the risk of bias (National Toxicology Program/Office for Health Assessment and Translation—NTP/OHAT) for the included studies. Finally, the data were synthesised narratively in accordance to synthesis without meta analyses (SWiM) method. **Results**: In total, 28 studies were included in this review. Most of the records (16) demonstrated increased residual DNA damage in radiosensitive individuals compared to normo-sensitive individuals based on γH2AX and TP53BP1. Overall, 15 studies included proteins other than DNA repair foci, of which five proteins were selected, Vascular endothelial growth factor (VEGF), Caspase 3, p16INK4A (Cyclin-dependent kinase inhibitor 2A, CDKN2A), Interleukin-6, and Interleukin-1β, that were connected to radiosensitivity in normal tissue and were reported at least in two independent studies. **Conclusions and implication of key findings**: A majority of studies used repair foci as a tool to predict radiosensitivity. However, its correlation to outcome parameters such as repair deficient cell lines and patients, as well as an association to moderate and severe clinical radiation reactions, still remain contradictory. When IR-induced proteins reported in at least two studies were considered, a protein network was discovered, which provides a direction for further studies to elucidate the mechanisms of radiosensitivity. Although the identification of only a few of the commonly reported proteins might raise a concern, this could be because (i) our eligibility criteria were strict and (ii) radiosensitivity is influenced by multiple factors. **Registration**: PROSPERO (CRD42020220064).

**Keywords:** ionizing radiation; normal tissue; biomarker; radiotherapy; radiosensitivity; proteomics

#### **1. Introduction**

#### *1.1. Background and Rationale*

Ionizing radiation is increasingly applied in medical therapy and diagnosis procedures. IARC Global Cancer Observatory reports more than 18 million new cases of cancer in 2018 (https://gco.iarc.fr/) [1] and radiotherapy (RT) is used to treat 50–60% of cancers [2].

**Citation:** Subedi, P.; Gomolka, M.; Moertl, S.; Dietz, A. Ionizing Radiation Protein Biomarkers in Normal Tissue and Their Correlation to Radiosensitivity: A Systematic Review. *J. Pers. Med.* **2021**, *11*, 140. https://doi.org/10.3390/jpm11020140

Academic Editors: Susan M. Bailey and Christophe Badie Received: 30 December 2020 Accepted: 14 February 2021 Published: 19 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

For medical imaging and image-guided interventions, the total exposure in the USA has increased 6-fold since 1980 [3]. However, potential adverse health effects of radiation exposure for patients, as well as for medical staff, especially with a focus on individual differences in radiosensitivity, are poorly understood.

Radiosensitivity is a measure for the reactions of cells, tissues, or individuals to ionizing radiation (IR). Subjects with increased reactions are described as radiosensitive, when compared to a majority of other "normal" responding individuals [4–6]. The reactions include inflammation, fibrosis, cardiovascular illness, cataracts, and cognitive decline [7]. The occurrence and severity varies among individuals and may be affected by genetic as well as by life style factors. In 5–10% of patients the use of RT is limited by the occurrence of acute, clinically diverse, strong radiogenic side effects of normal tissue in the radiation field, leading to suboptimal tumour control or to serious impairment of the quality of life for patients [8–10]. A reliable, pre-therapeutic identification of radiosensitive patients would improve therapy because an individual dose adjustment could be applied. Furthermore, the identification of radiosensitive persons would be a valuable step in the protection of occupationally exposed persons. To foster research in this field, two radiation research platforms, Multidisciplinary European Low Dose Initiative (MELODI) and European Alliance Medical Radiation Protection Research (EURAMED), declared individual differences in radiation sensitivity as a key research priority.

In a small subset of patients the severe reactions can be ascribed to known radiation hypersensitivity syndromes, such as Ataxia–Telangiectasia (A–T), Fanconi anaemia (FA) or Nijmegen Breakage Syndrome (NBS) [11–13]. As late as 2010, children with A–T mutations have succumbed to death following RT [14]. These genetic syndromes, however, only comprise about 1% of the patients demonstrating severe side effects [15] and most of the enhanced tissue reactions cannot be explained by known genetic disorders.

Some further genetic associations were suggested by candidate gene approaches as well as by genome-wide association studies in radiotherapy patients. However, only a small proportion of radiosensitive individuals could be identified [16]. Additionally, functional assays such as DNA double stand break repair, induction of chromosomal aberrations, and radiation-induced apoptosis in ex vivo irradiated blood lymphocytes, have been described as predictors of radiosensitivity [17]. In parallel, a substantial number of IR-induced transcriptional and translational alterations were reported [18]. These studies benefit from recent technical developments in omics applications, which facilitate the cost effective quantification of numerous candidates, including posttranslational modifications of proteins. However, for most of the candidates, the potential correlation between IRinduced deregulation and radiosensitivity is under discussion.

Therefore, the purpose of this paper is to present the *status quo* of our knowledge of IR-induced changes in protein expression in normal tissue that can be correlated to radiosensitivity. We focus on proteins and protein modifications, as, due to posttranscriptional regulatory processes, the alterations in protein levels may describe the actual cell state, inclusive stress responses, more precisely than transcriptome changes [19]. The future goal will then be to establish protein biomarkers for the identification of radiosensitive or radio-resistant individuals. This will help to personalise treatment strategies to cancer patients during RT or help to assist an individualised risk assessment process by identifying and protecting occupationally radiation-exposed persons.

#### *1.2. Objectives*

The aim of this systematic review (SR) is to investigate the IR-induced changes, both in vivo and in vitro, in the human proteome that can be correlated to radiosensitivity.

#### **2. Methods**

#### *2.1. Protocol and Registration*

The review protocol [20] was registered to International Prospective Register of Systematic Reviews (PROSPERO) on 10.11.2020 (CRD42020220064).

#### *2.2. Eligibility Criteria*

Studies that comply with elements of Population, Exposure, Comparators, and Outcome (PECO) were eligible for this SR. The full description of PECO parameters was provided in the protocol [20]. In short, the population for this SR were primarily humans or human-derived non-tumour tissue and secondary non-tumour cell lines that were exposed to ionizing radiation. This population was compared to non-exposed individuals or *in vitro* cultures. Changes in expression of proteins after the exposure, which were associated with radiosensitivity, were defined as the outcome of this review. Only primary peer-reviewed published studies in English language were considered. As a study on ionizing radiation protein biomarkers for epidemiological studies was published in 2012 [21], studies between 2011 and 2020 were investigated in this SR.

#### *2.3. Information Sources*

Studies were identified using electronic databases and with consultations of field experts. The authors of the studies were not contacted for further studies or questions regarding the paper.

#### *2.4. Search*

NCBI PubMed (https://pubmed.ncbi.nlm.nih.gov/) [22] and ISI Web of Knowledge (v.5.34) (https://www.webofknowledge.com/) [23] were used to perform the searches. In addition, papers were also added manually. Search strings included a combination of population, exposure, and outcome elements and the applied search strings for ISI Web of Knowledge are provided in Supplementary Information 1. The Pubmed IDs of identified studies from manual as well as database searches were entered in Microsoft Excel and the duplicates (same studies in different databases) were removed using the built-in "Remove duplicate" tool.

#### *2.5. Study Selection*

A two-phase screening was performed by authors Dietz and Subedi in parallel. In phase I screening, title, abstract, and key words of all of the studies were cross-checked manually with the inclusion and exclusion criteria provided in the protocol [20]. The articles that were excluded after phase I screening are provided in Supplementary Information 2. A phase II screening (full-text screening) was performed on the remaining articles after phase I screening. The articles excluded after phase II screening, along with the reasons excluded are also given in Supplementary Information 2. Any disagreements between the reviewers was solved either in consensus, or by involving a third reviewer (Moertl or Gomolka) if necessary. The articles retained after phase II screening were used for Synthesis without Meta-analyses (SWiM).

#### *2.6. Data Collection Process*

The data collection was performed in Google Sheets by Subedi, Dietz, and Moertl, with one reviewer entering the data and the other person confirming it. The data were finally processed with Microsoft Excel. The form for data extraction is submitted in Supplementary Information 3, along with this review. Any disagreements were solved by consensus or by involving a third reviewer. In the case of missing information, the authors were not contacted and was denoted with 'nr'.

We extracted information about: the name of the protein; the fold change ratio after IR; bio fluids or cell lines being investigated; the method used to determine the fold change; the quality and quantity of IR; the characteristics of the donor(s) (age, sex, and diseased or healthy); eligibility criteria of the patients; the method used to quantify radiosensitivity (e.g., viability testing); the replicates performed for the experiment and the statistics to accompany the fold changes; the outcome of the change in protein expression; posttranslational modification; and conflict of interest. The findings were summarised and the heterogeneity of the data was compared visually in form of tables.

#### *2.7. Grouping Studies for Synthesis*

This SR was performed to investigate the changes in protein expression in normal tissue after exposure to ionizing radiation. Therefore, the in vivo and in vitro studies were grouped together and no differences were made between the different radiation qualities. The doses are provided in Gray (Gy) and the dose-rates are provided in (Gy/min).

#### *2.8. Standardised Metric and Transformation Used*

The increase or decrease in protein expression after IR (fold changes, Equation (1)) was used as a measure of effect size of the exposure. The fold changes were not calculated in this manuscript but taken from the respective studies.

$$\text{Fold change} \,(\text{protein}) = \frac{\text{Protein expression after IR}}{\text{Protein expression before IR}} \tag{1}$$

#### *2.9. Synthesis Methods*

For each comparison, the null hypothesis represented by *p*-value, or in certain cases by an adjusted *p*-value resulting from multiple testing, was used as synthesis method for each outcome.

#### *2.10. Certainity of Evidence*

Studies which contained commonly deregulated proteins were pooled together. Studies were given an initial confidence rating of high, moderate, low, or very low based on the presence of features (controlled exposure, exposure prior to outcome, individual outcome data, and the use of comparison group). Following the OHAT method, which is based on Grading of Recommendations Assessment, Development and Evaluation (GRADE) working group guidelines, the studies were up- or downgraded. The factors increasing confidence were magnitude of the effect, dose response, residual confounding, and consistency, whereas the factors decreasing confidence were risk of bias, unexplained inconsistency, indirectness, and imprecision.

#### **3. Results**

After database searching and inclusion of manual sources, 2733 studies were identified. The records were screened for title, abstract, and key words, and 100 articles were selected for a full-text review. Finally, 28 articles were included for this SR (Figure 1). In the included articles, 13 studies examined solely DNA repair foci, 12 studies investigated proteins other than repair foci, with 3 studies also including repair foci.

#### *3.1. Study Characteristics of the included Articles*

The 16 studies that used repair foci to determine individual differences in radiosensitivity included 10 cohort studies (van Oorschot et al., 2014 [24], Vasireddy et al., 2010 [25], Bourton et al., 2011 [26], Mumbrekar et al., 2014 [27], Poulilou et al., 2015 [28], Lobachevsky et al., 2016 [29], Buchbinder et al., 2016 [30], Granzotto et al., 2016 [31], Djuzenova et al., 2013 [32], and Goutham et al., 2012 [33]) and 6 model system (Vandersickel et al., 2010 [34], Martin et al., 2014 [35], Martin et al., 2011, [36], Minafra et al., 2015 [37], Miyake et al., 2019 [38], and Nguyen et al., 2019 [39]). The detailed study characteristics of these studies is provided in Table 1a.

(28).

**Figure 1.** PRISMA flowchart that displays the number of records identified (2733), the number of records screened for a full-text review (100), and the number of records included in the review **Figure 1.** PRISMA flowchart that displays the number of records identified (2733), the number of records screened for a full-text review (100), and the number of records included in the review (28).

*3.1. Study Characteristics of the included Articles* The 16 studies that used repair foci to determine individual differences in radiosensitivity included 10 cohort studies (van Oorschot et al., 2014 [24], Vasireddy et al., 2010 [25], Bourton et al., 2011 [26], Mumbrekar et al., 2014 [27], Poulilou et al., 2015 [28], Lobachevsky et al., 2016 [29], Buchbinder et al., 2016 [30], Granzotto et al., 2016 [31], Djuzenova et al., 2013 [32], and Goutham et al., 2012 [33]) and 6 model system (Vandersickel et al., 2010 [34], Martin et al., 2014 [35], Martin et al., 2011, [36], Minafra et al., 2015 [37], Miyake et al., 2019 [38], and Nguyen et al., 2019 [39]). The detailed study characteristics of these studies is provided in Table 1a. Amongst the studies, which investigated proteins other than repair foci, 15 studies were included: five cohort studies (Braicu et al., 2014 [40], Rodruiguez-Gil et al., 2014 [41], Skiöld et al., 2015 [42], Yu et al., 2018 [43], and Lacombe et al., 2019 [44]) and 10 studies on model systems (Cao et al., 2011 [45], Park et al., 2012 [46], Fekete et al., 2015 [47], Minafra et al., 2015 [37], Nishad and Ghosh, 2015 [48], Shimura et al., 2015 [49], Yim et al., 2017 [50], Miyake et al., 2019 [38], Nguyen et al., 2019 [39], Moertl at al., 2020 [51]). In total, 5 of these 10 studies were conducted with peripheral blood mononuclear cells (PBMCs) (Yu et al., 2018, Nguyen et al., 2019, Lacombe et al., 2019, Skiöld et al., 2015, and Nishad and Ghosh, 2015), and one with PBMCs-derived extracellular vesicles (Moertl et al., 2020). The detailed study characteristics are provided in Table 1b.

Amongst the studies, which investigated proteins other than repair foci, 15 studies were included: five cohort studies (Braicu et al., 2014 [40], Rodruiguez-Gil et al., 2014 [41], Skiöld et al., 2015 [42], Yu et al., 2018 [43], and Lacombe et al., 2019 [44]) and 10 studies on model systems (Cao et al., 2011 [45], Park et al., 2012 [46], Fekete et al., 2015 [47], Minafra et al., 2015 [37], Nishad and Ghosh, 2015 [48], Shimura et al., 2015 [49], Yim et al., 2017 [50], Miyake et al., 2019 [38], Nguyen et al., 2019 [39] , Moertl at al., 2020 [51]). In total, 5 of these 10 studies were conducted with peripheral blood mononuclear cells (PBMCs) (Yu In total, the 28 included studies identified 76 proteins, which were correlated with normal tissue radiosensitivity. The results were prioritised so that the proteins identified in more than one study, regardless of the direction of regulation, along with their role in radiation response, were described further. Besides changes in repair foci (γH2AX and TP53BP1 quantities), the proteins were identified in more than one study are Vascular endothelial growth factor (VEGF), Caspase 3, p16INK4A (Cyclin-dependent kinase inhibitor 2A, CDKN2A), Interleukin-6, and Interleukin-1B.

et al., 2018, Nguyen et al., 2019, Lacombe et al., 2019, Skiöld et al., 2015, and Nishad and Ghosh, 2015), and one with PBMCs-derived extracellular vesicles (Moertl et al., 2020). The

detailed study characteristics are provided in Table 1b.





96






#### *3.2. IR-Induced Changes in Repair Foci Proteins*

H2AX, a variant of the histone protein H2A, is located in the nucleus and its functions include chromatin organisation and DNA damage response. In case of DNA double strand break damage, its phosphorylation by PI3 kinases ATM, ATR, and DNAPKcs signals the damaged site, and recruits downstream DNA repair proteins [52–55]. The phosphorylated isoform on serine 139 is termed as γH2AX [52,53]. The initial γH2AX signal develops and expands within the first hour after DNA damage induction. With subsequent repair of the damaged sites, the signal decreases again. Depending on the amount and the complexity of the DNA damage and on DNA repair capacity, the differences in DNA repair kinetic and residual foci level are observed [56]. In addition to γH2AX, another component of the DNA double strand break repair machinery, TP53BP1 (Tumour Protein P53 Binding Protein 1) [32,36], was also identified as a target candidate to predict radiation sensitivity. TP53BP1 plays an essential role in the canonical non-homologous end joining (NHEJ) repair of DNA double strand breaks (DSB), which is the main repair pathway of DSB in G0–G1 cell cycle phase, e.g., in peripheral blood lymphocytes [57]. TP53BP1 clusters appears during radiation response and disappears in a similar time dependent kinetic as γH2AX foci do. γH2AX and TP53BP1 quantities were measured by immunofluorescence microscopy in most of the studies except for Bourton et al. and Pouliliou et al. In their studies, γH2AX expression was analysed by fluorescence-activated cell sorting (FACS) and western blot, respectively. The IR-induced alterations of γH2AX and TP53BP1 expressions are presented in detail in Supplementary Information 4.

In all studies, irradiation was performed with gamma or X-ray radiation at a high dose rate and doses from 0.5 2.0 Gy. Studies were performed in different cell lines (fibroblast, lymphoblastoid, epithelial cell lines) harbouring DNA repair defects, or in primary cells (blood cells, hair follicle) from cancer patients. From all parameters investigated, such as basal foci level, radiation induced foci and residual foci at later repair time points, elevated levels of residual γH2AX or TP53BP1 foci appear to be robust to identify radiosensitive cells or individuals.

DNA repair deficient individuals demonstrate delayed development of the initial DNA damage or delayed DNA repair, resulting in an increased level of residual damage after 24 hours [35,36]. Therefore γH2AX is considered as a putative predictive biomarker to detect radiation sensitive individuals harbouring DNA repair defects by performing an *in vitro* challenging assay and investigating signal development and disappearance [26,56,58]. Promising studies demonstrating a positive association of increased residual damage in ATM [35,36,59], Ligase IV deficient radiation sensitive individuals [30,36], and in cancer patients experiencing strong acute or late side effects from the radiation treatment [24–29,32] are presented. However, the literature overview has shown multiple factors, such as high variability of the assay itself, the lack of a standardized protocol including a fixed in vitro exposure dose, repair time point to analyse residual foci, and comparator group, bias the results. Therefore, correlation to outcome parameters such as genetically defined repair deficient cell lines and patients, as well as association to clinical radiation sensitivity, still remain contradictory [5,34,60]. Our systematic review and others show that although γH2AX and TP53BP1 expressions have the potential to predict an in vitro radiation response in a number of patients; large cohorts need to be analysed by standardised protocols to improve the robustness and sensitivity of the assay, and to decipher the subgroups of patients for which the assay is a meaningful tool to predict detrimental radiation reactions [5,18,61].

#### *3.3. IR-Induced Deregulated Proteins Excluding Repair Foci and Risk of Biases*

An aim of our SR is to discover new feasible markers on protein levels that are associated with radiosensitivity, besides repair foci proteins. To provide a rich reflection of evidence for the reader, we included both significantly deregulated and not deregulated proteins in Supplementary Information 5. There is comparatively little evidence published on this topic within the inclusion parameters specified (especially the correlation to radiosensitivity). Therefore, if the studies included experiments that depict cell survival, the

paper was incorporated to the synthesis, irrespective of a direct correlation of the outcome to radiosensitivity. Table 2 presents the evaluation of studies containing proteins, other than repair foci, on all applicable risk of bias (RoB) questions as developed by the Office of Health Assessment and Translation (OHAT) [62]. The questions concerning the RoB tools and the criteria to judge the different biases are provided in the protocol [20] and in Supplementary Information 6. Although a set of 11 questions was used to evaluate the studies, the studies were categorised into three tiers (T1, T2, or T3) primarily based on the responses to the following key questions (Supplementary Information 7)


None of the studies were categorised in T1, one study (Park, 2012) was categorised into T3, and the rest were categorised as T2. The RoB questions are suited to cohort and human clinical trials compared to model systems. Concealment, randomisation, and blinding in most studies on model systems are not performed because (i) it is usually a single person that performs the studies and (ii) it is not a common practice to conceal the study groups from the researcher. Therefore, most studies received a 'probably high risk of bias' assessment in randomisation, concealment, and blinding domains. Although randomisation is performed during the accessing of outcomes, for example, when performing mass spectrometric analyses or measuring γH2AX quantities on coded slides, more often than not, it is not reported to ensure brevity during publication. Based on the results from this SR, we can recommend that studies on model systems should take care of randomisation, concealing of study groups, blinding the accessors, and, most important, reporting them.


**Table 2.** Accessing risk of bias for studies that included proteins other than repair foci.

The proteins that were reported in at least two studies (Table 3) are explained further:


**Table 3.** List of proteins identified in at least two studies, not concerning repair foci.
