RRM1 gene expression evaluated in the liquid biopsy (blood cfRNA) as a non-invasive,predictive factor for radiotherapy-induced oral mucositis and potential prognostic biomarker in head and neck cancer patients

1. Introduction

Head and neck cancers (HNC) with about 650 thousand new cases a year are the sixth most prevalent group of malignant neoplasms in the world [16, 30]. Radiotherapy (RTH) is a recognized method of radical treatment of patients with HNC. RTH instead of surgery or as complementary treatment it lets to achieve 75–95% cures with maintaining the organ function, especially in early stages of disease progression (Grade I–III). Standalone RTH with altered fractioning of radiation and its combination with cytotoxic and/or molecularly targeted drugs, with appropriate selection of patients, contribute to significant increase of efficiency of this treatment. Unfortunately, intensified treatment, it is also associated with a series of adverse effects including oral mucositis (OM) occurring, to a various degree, in almost all patients [27, 28].

OM is described as a gradually increasing edema of mucous membranes accompanied by oral erythema, ulceration and often also pain and swallowing disorder. In almost 1/3 patients there occurs severe OM (of 3 rd and 4 th degree in RTOG/EORTC scale) [3, 14, 27]. Severe OM leads to interruptions in RTH, necessity of frequent hospitalizations as well as deterioration of the quality of patients’ life [3]. In 2/3 patients it is necessary to reduce the doses of cytostatics and, in other patients, delay or discontinuation of treatment must be considered [33]. It is associated with poorer treatment efficiency (lower rates of local control), higher risk of progression and shorter overall survival (OS) [3, 21, 25].

Despite the known OM risk factors such as male gender, old age, lack of or poor oral hygiene, high dose of radiation, smoking, systemic diseases, RTH technique and combined radiochemotherapy (RCTH), so far no factors have been identified which would facilitate precise estimation of the risk of occurrence and intensity of OM [24, 25]. Due to significant individual variability in the development of OM in patients characterized with the same demographic and clinical features, the key role in the modification of the risk of disease progression may be played by genetic factors [4, 7].

Ribonucleotide reductase (RNR) – a reductase group enzyme participating in the synthesis of deoxyribonucleic acids. It facilitates the acquisition of de novo deoxyribonucleoside diphosphates from ribonucleoside diphosphates. Apart from salvage pathway, in which nucleotides are retrieved through phosphorylation of nucleosides, it is the only process of acquiring endogenous deoxyribonucleoside triphosphates – necessary for the synthesis of DNA. A gene coding for RNR1 subunit of the enzyme (RRM1) is located on chromosome 11p15.5. Total loss of RRM1 gene is lethal whereas partial loss of the function is normally sufficient to initiate the neoplastic process [17, 18].

The predictive and prognostic value of RRM1 gene has been intensively studied in advanced non-small cell lung cancer (NSCLC), in which it was observed that the response to gemcitabine based chemotherapy (CTH) and survival are strongly correlated with the level of this gene expression [18, 23, 29, 32].

Ionizing radiation causes structural damage of the DNA helix. It results in cell death, most frequently by apoptosis. DNA damage is recognized by specialized proteins of the numerous DNA repair mechanisms. In case of disorders including double strand breaks, they are mainly: NHEJ (non-homologous end-joining) and HR (homologous recombination), and in case of single strand breaks: NER (nucleotide excision repair) and BER (base excision repair) [5].

It seems that among many factors associated with the development of OM induced by ionizing radiation or/and CTH, the decisive role is played by proteins involved in DNA repair. The disturbance of correct functioning of proteins crucial for the synthesis and supply of nucleotides including RNR may impair the efficiency of cellular mechanisms of DNA repair. Nucleic acid damage caused by RTH or RCTH is repaired by the systems of DNA repair dependent on the activity of RNR [17].

The objective of this study was to evaluate the correlation between expression of RRM1 gene measured in free circulating RNA (cfRNA) and the risk of more severe OM and disease-free survival (DFS) and overall survival (OS) in patients undergoing RTH for HNC.

Radiation-induced mucosa reactions (grade 0–3 according to RTOG) developed in all study participants during subsequent cycles of RTH treatment. We observed intensification of RTH side effects after each week of radiation. After the first week of RTH, most patients demonstrated lack or mild toxicity of applied treatment (grade 0 in 8.3% of patients and grade 1 in 91.7% of patients). Mucositis toxicity increased after successive weeks of RTH (after 3 rd week, grade $>$ 2 of acute mucositis reaction was observed in 10% of patients, whereas after 7 th week in 45% of subjects). Generally patients’ data were not correlated with either the severity of RTH toxicity or the relative risk of OM development however two exceptions were noted: smoking history and prior alcohol consumption (Table 2). In former smokers compared to those currently smoking significantly higher risk of 3 grade of OM after 5, 6 (the same results: OR = 8.22; 95% CI [1.461–46.723], p = 0.017) and 7-th week of RTH was found (OR = 19; 95% CI [2.881–125.317], p = 0.002). Occasional alcohol consumption was associated with lower risk of more serve OM (grade $>$ 2), however this relationship was noted only after 7-tf RTH cycle (OR = 0.15; 95% CI [0.037–0.643], p = 0.01).

The level of RRM1 gene expression was equally distributed in the study group, 50% HNC patients had high and 50% had low expression. Expression did not vary between patients with different clinical-demographic data including age, gender or treatment features (Table 3). Moreover, the level of RRM1 gene expression significantly differed between patients with mild and acute mucositis. These differences were observed after each week of RTH, excluding week one. In all cases, patients with grade 3 of OM had higher gene expression compared to individuals with OM grades 1 and 2. Differences in gene expression after subsequent weeks of RTH in patients with different grade of OM are presented in Fig.1. We also noted that after weeks 5, 6 and 7 of RTH, patients with severe, grade 3 toxicity had significantly higher RRM1 gene expression compared to patients with less severe OM – grades 1 and 2 (week 5: p = 0.020; week 6: p = 0.039; week 7: p = 0.037) (Table 4). Interestingly, RRM1 gene level had significant impact on grade 3 risk of OM development after treatment weeks 5, 6 and 7. HNC patients with high gene expression had significantly higher risk of development of grade 3 OM after 5 th week of RTH (OR = 4.971; 95% CI [1.387–17.816], p = 0.014). Also in patients with high gene expression, grade 3 of OM developed more frequently compared to patients with low RRM1 gene level (OR = 4.333; 95% CI [1.203–15.606], p = 0.025). After 7th cycle of therapy individuals with high gene expression had higher risk of development of grade 3 toxicity compared to patients with low gene level (OR = 3.50; 95% CI [1.201–10.196], p = 0.022). Impact of RRM1 gene expression on the risk of development of OM is presented in Table 4. To investigate whether similar relationships exist in patients treated with RTH alone, we have analyzed this subgroup ( n = 38). It was found that, patients with high RRM1 gene expression had significantly higher risk of 3 grade of OM after 5, 6 as well as 7 cycle of RTH (Table 5).

Considering gene expression prior to RTH and grade of OM during subsequent weeks of treatment, we used ROC analysis to set an expression cut-off point to diagnose grade 3 OM. The analysis was mostly useful in the diagnosis of grade 3 OM after RTH weeks 5, 6 and 7. In week 5, the diagnostic accuracy was: 63.6% sensitivity and 77.3% specificity (AUC = 0.719 [0.582–0.832]), in week 6 it was: 84.6% and 60.5% (AUC = 0.722 [0.586–0.833]), and in week 7: 72.7% and 63.6% (AUC = 0.735 [0.599–0.845]) (Fig. 2A, B and C).

Based on available data on DFS and OS we performed treatment outcome analysis. Low expression of RRM1 gene was not significantly related with higher risk of DFS shortening (18 months vs 24 months; HR = 1.65; 95% CI [0.490–5.578], p = 0.446; Fig. 3A). Similarly, low expression of RRM1 gene was not significantly related with higher risk of OS shortening (22 months vs 34 months; HR = 1.97; 95% CI [0.597–6.521], p = 0.294; Fig. 3B).

OPEN IN VIEWER

OPEN IN VIEWER

Ionizing radiation disrupts the structure of DNA leading to the occurrence of specific changes, i.e. double-strand breaks (DSB), single-strand breaks (SSB) and even base impairment, which may lead to halting of the cell cycle and then apoptosis. DNA Damage Response (DDR) plays a key role in the biological effects of ionizing radiation. Cellular response to RTH is largely associated with direct DNA damage, mostly in the form of DSB. However, paradoxically, only a few studies concentrated on SNPs located in genes, whose protein products are mostly involved in DSB repair. On the contrary, the researchers mainly concentrated on ERCC and XRCC families of genes involved in the mechanisms of SSB damage repair (NER and BER respectively). Interestingly, cellular model studies (fission yeast) demonstrated that RNR is one of the enzymes necessary for correct performance of HR repair of DNA damaged in DSB mechanism [2, 5, 13, 35].

Since RRM1 gene is located on chromosome 11p15.5 and the loss of heterogeneity (LOH) of this region was commonly observed in a series of neoplasms some researchers have suggested that this gene has tumor suppressor properties. It is confirmed by the fact that total loss of RRM1 is lethal (due to total blockage of RNR function) and partial loss of the function is normally sufficient to initiate the neoplastic process [17, 18]. However, it should be also taken into consideration that LOH in the region containing RRM1 gene as well as loss of other regions involving candidate gene polymorphisms of DNA repair pathways may be mainly related with an inherited susceptibility to cancer. Xu et al. found that A/A genotype of polymorphism RRM1gene (*151A>T) as well as the T/T + T/C genotype of RRM1 (-756T>C) was positively associated with an increased risk of lung cancer. Moreover, the T/T+G/T genotype of RRM1 gene -585T>G was associated with a decreased risk for tumorigenesis, suggesting that these SNPs may have potential function to influence expression and amplification of RRM1 gene [39]. However, despite, that none of the SNPs located within RRM1 gene has been linked to head and neck cancer susceptibility yet, many examples of that kind of relationship has been found in the case of other genes involved in different DNA repair pathways [20].

Results of several studies have been published so far regarding radiosensitivity and radioresistance mediated by molecular changes in HNC patients, however, only few of them analyzed the occurrence of toxicity of RTH. They were almost exclusive studies on polymorphisms (SNPs) of genes associated with processes underlying the development or response to RTH (whose protein products are involved in: DNA repair mechanisms, inflammatory processes, apoptosis, signal pathways and reduction of ROS influence) [2, 40].

Pratesi et al., in a study of patients irradiated due to HNSCC, demonstrated a correlation between SNPs of XRCC1 as well as RAD51 genes with sever OM and accompanying dysphagia [26]. In turn, Kornguth et al. observed, in patients with HNC, a correlation between faster healing of lesions caused by RTH with SNP of ERCC4 gene [6]. Also in case of neoplasms other than HNC, a significant correlation was showed between genetic changes within genes associated with DNA repair and development of acute radiation reaction [1].

RR status is a well-known factor modulating CTH efficiency, which mostly applies to schemes based on gemcitabine and to a certain extent also on 5-fluorouracil because protein product of RRM1 gene is a molecular target of the above mentioned cytostatics. Gemcitabine is characterized with high activity against nasopharyngeal cancer. It was demonstrated that high level and/or mutations/SNPs in RRM1 gene are associated with the occurrence of resistance to gemcitabine [18, 23, 29, 32].

In the study Rodriguez and coworkers performed in colorectal cancer (CRC) patients treated with gemcitabine-based chemotherapy RRM1 gene polymorphism (rs12806698) was associated with clinical efficacy (longer time to progression in CC genotype carriers). Moreover in vitro experiments, in CRC cell lines, showed that the RRM1 gene genotype was associated with the levels of RRM1 protein expression and with gemcitabine IC50 values. In their experiments the down-regulation of RRM1 with a specific siRNA strongly influenced gemcitabine sensitivity [12]. Moreover, the recently published study revealed also a significant correlation of SNP (rs9937) of RRM1 gene with lower survival rates in the cytarabine-based induction treatment in acute myeloid leukemia patients [15]. In paper Lee et al. SNP (rs9937) of RRM1 gene and haplotypes ATAA and ATGA were significantly associated with neurotoxicity in patients with metastatic breast cancer receiving gemcitabine plus paclitaxel chemotherapy [34]. In the study, Zeng et al. patients with pancreatic cancer with genotype AA of RRM1 gene SNP (rs1662172) treated with radiation, had a significantly higher risk of shorter survival [10]. Feng et al. found that in the case of SNP (rs11030918) of the RRM1 gene, the carriers of the CC genotype had a significantly lower chance of responding to the treatment according to schemes based on platinum compounds. In contrast, Kim et al. showed a higher response rate to gemcitabine-based chemotherapy regimens in patients with the AC/CT allelotype (rs12806698/rs11030918) of the RRM1 gene. Moreover Dong et al. showed significant differences in the PFS depending on (rs12806698) SNP. Also in our previous publication the median PFS was significantly longer in carriers of AA (rs12806698) as well as CC rs11030918) genotype of RRM1 gene compared to patients with other genotypes. In addition, the CC genotype carriers (-37C>A) showed a significant increase in the risk of shortening overall survival (OS) in comparison to patients with AA or AC genotypes [29]. One of few studies that evaluated the predictive value of determination of the level of RRM1 gene mRNA expression in peripheral blood (as well as corresponding cancer tissue) was described in a publication by Zhang et al. [9]. It involved patients with advanced non-squamous cell lung cancer treated with a combination of cisplatin and gemcitabine. They demonstrated that low expression of RRM1 gene both in peripheral blood and in neoplastic tissue was associated with better response to CTH, longer median PFS and OS. Moreover it turned out that there is a statistically significant positive correlation between the level of RRM1 gene expression in peripheral blood and neoplastic tissue [9].

Hitherto, the analysis of expression of mRNA of genes important for HNC treatment efficiency evaluation (ERCC1 or RRM1) was performed almost exclusively in the tumor tissue. However, the material obtained in this way is mostly used up in the diagnostic process and often there is not enough for performing additional molecular examinations. Moreover, dynamic changes occurring in tumor in course of the applied therapy may require analysis of freshly collected material at a given stage (second biopsy), which, for many reasons, is usually not performed. It is necessary to look for a simple and convenient method of evaluation of selected biomarkers facilitating treatment individualization [9]. Vogel et al. demonstrated, in patients with NSCLC, a correlation between DNA repair capacity (DRC) and expression of ERCC1 [36]. Dong et al. successfully extracted mRNA from mononuclear peripheral blood cells (PBMC) and the mRNA was found to be suitable for RRM1 gene expression evaluation. However they found no significant correlation between expression of RRM1 gene and the gemcitabine treatment outcome [32].

RNA circulating in serum was successfully isolated in a series of neoplasms including: lung, nosopharyngeal, pancreatic, renal and melanoma. Nucleic acids may be actively released to circulation by neoplastic cells in the form of microvesicles (which additionally protects them from RNases) [8, 37]. It seems that the application of the evaluation of cfRNA instead of cfDNA both in diagnosing and monitoring of the ongoing neoplastic process may be more beneficial for several reasons. The level of mRNA in normal cells is strictly controlled while neoplastic progression is usually associated with progressive deregulation of expression of many genes. Contrary to DNA, which, in the context of the changes, is the permanent molecule (generation and consolidation of new mutations is a long term process), the changes in gene expression, reflected in the level of mRNA, may be sudden (taking days or even hours). What is more, the level of cfRNA in cancer patients is often higher than of cfDNA [8, 37]. Therefore, the changes in the mRNA level seem to be a more sensitive biomarker of changes occurring in a neoplasm in response to the applied treatment.

The level of RRM1 gene cfRNA may depend on a number of factors, including its source (it may reflect various expression levels in different tissue types – with highest levels in skin and lymph nodes) and complex regulation of expression especially in relation to the various cancer type [19, 22]. However, the possible mechanism of altered level of expression (in solid tissue and finally in circulation) in some patients may be associated with the genetic polymorphisms (e.g. SNPs in the gene regulatory region) that influence transcript level [31]. Few such SNPs has been recently reported [11, 38].

Although we have failed to demonstrate a statistically significant association between the RRM1 expression level and DFS and OS (data regarding survival is still immature), the Kaplan-Meier graphs show, that curves (for both cases) were well separated which may suggest a trend for significant results (probably study conducted on a larger group of patients is needed).

It seems, that modalities of RTH as well as the occurrence of adverse effects may be diversified not only in relation to tumor type and location, the degree of progression, patient habits, but also to the occurrence of specific molecular phenotype of the tumor.

5. Conclusion

The evaluation of the level of RRM1 gene expression in RNA isolated from peripheral blood plasma allows for estimation of the risk of development of severe OM in HNC patients treated with RTH. The evaluation of cfRNA (alongside cfDNA) may in near future become a promising tool of noninvasive diagnostics helpful in early detection and monitoring of the ongoing neoplastic process. This type of liquid biopsy can also serve to detect relapse, forecast the course of the disease, response to treatment and patient survival.