Assessing patient characteristics and radiation-induced non-targeted effects in vivo for high dose-rate (HDR) brachytherapy.

PURPOSE
To test whether blood, urine, and tissue based colony-forming assays are a useful clinical detection tool for assessing fractionated treatment responses and non-targeted radiation effects in bystander cells.


MATERIALS AND METHODS
To assess patients' responses to radiation treatments, blood serum, urine, and an esophagus explant-based in vivo colony-forming assay were used from oesophageal carcinoma patients. These patients underwent three fractions of high dose rate (HDR) intraluminal brachytherapy (ILBT).


RESULTS
Human keratinocyte reporters exposed to blood sera taken after the third fraction of brachytherapy had a significant increase in cloning efficiency compared to baseline samples (p < 0.001). Such results may suggest an induced radioresistance response in bystander cells. The data also revealed a clear inverse dose-rate effect during late treatment fractions for the blood sera data only. Patient characteristics such as gender had no statistically significant effect (p > 0.05). Large variability was observed among the patients' tissue samples, these colony-forming assays showed no significant changes throughout fractionated brachytherapy (p > 0.05).


CONCLUSION
Large inter-patient variability was found in the urine and tissue based assays, so these techniques were discontinued. However, the simple blood-based assay had much less variability. This technique may have future applications as a biological dosimeter to predict treatment outcome and assess non-targeted radiation effects.


Introduction
High dose-rate intraluminal brachytherapy (HDR-ILBT) has established itself as an eff ective treatment modality for patients diagnosed with advanced stages of esophageal cancer (Sur et al. 2002). Brachytherapy enables high doses of radiation to be delivered to the tumor to improve the patients ' dysphagia scores, quality of life (Berry et al. 1989), has the advantage of providing conveniently fast outpatient procedures (Sur et al. 1998). A remote afterloading HDR unit is used for delivering high doses of gamma radiation, from an  ) source, to the tumor site (Sur et al. 2002). A rapid drop-off of dose from the treatment site to surrounding normal tissues results in a very small risk of injury to nearby normal tissues (Yoshioka et al. 2013). Fractionated HDR-ILBT has shown to signifi cantly improve dysphagiafree survival and longevity in comparison to other palliative modalities (Sur et al. 1998).
Dose to normal tissues typically restricts treatment planning protocols for radiotherapy modalities, as these are limited by normal tissues tolerance doses (Mothersill et al. 2004a). However, in vitro research has documented the biological implications of bystander factors being released into non-irradiated cells which has been shown in the literature to trigger a cell death response (Mothersill et al. 2004b). Consequently, non-targeted radiation eff ects can ultimately aff ect treatment planning protocols, as there is a possibility of much larger out-of-fi eld eff ects in normal tissues than initially expected (Butterworth et al. 2013). Other work has suggested that radiation-induced bystander eff ects (RIBE) may provide insight into understanding the effi cacy of radiotherapy, as bystander factors may enhance tumor cell killing (Boyd et al. 2008, Prise and O ' Sullivan 2009, Butterworth et al. 2013. Currently, it is not fully understood whether the release of bystander signals into healthy surrounding tissues, near radiation fi elds, leads to unwanted damage in normal cells (Brenner et al. 2000, Hall and Wuu 2003, Boyd et al. 2008. Th erefore we and others have extended the investigation of non-targeted radiation eff ects from an in vitro experimental approach (Mothersill and Seymour 1997, Prise et al. 1998, Lyng et al. 2000 to the whole organism by using animal (Morgan 2003, Chai and Hei 2008, Koturbash et al. 2008) and human models , Mothersill et al. 2002, Marozik et al. 2007, Chai and Hei 2008.
For the past several years, non-targeted radiation eff ects such as clastogenic eff ects Mothersill 2006, Howe et al. 2009), RIBE (Emerit et al. 1995, Mothersill and Seymour 1997, Ryan et al. 2009) and adaptive responses have been well documented. Radiation-induced clastogenic eff ects are found in atomic bomb survivors (Pant and Kamada 1977), humans undergoing radiotherapy , and in the blood serum collected from Chernobyl liquidators (Marozik et al. 2007).
One of the earlier studies observing clastogenic eff ects was published by Goh and Sumner (1968); the study evaluated chromosomal aberrations in cultivated leukocytes treated with blood plasma taken from patients that underwent total body irradiation. Th e fi ndings showed that blood plasma exposed to radiation increased the number of chromosome breaks in leukocytes compared to unirradiated samples. Similar clastogenic eff ects have also been reported by other investigators in the literature with humans (Hollowell and Littlefi eld 1968, Pant and Kamada 1977, Emerit et al. 1994, 1995 and animals (Faguet et al. 1984) following radiation exposure.
Earlier work focused on exploring the variability inherent in human urothelial tissue explants and their ability to express bystander signals in reporter cells (Mothersill et al. 2002). Signal production was found to be sex-specifi c and had a dependence on whether the participants had no existing malignancies. A gender discrepancy was observed, tissue samples harvested from female participants resulted in a higher reduction in cloning effi ciency compared to males. Other researchers used a rodent model to assess nontargeted radiation eff ects within non-irradiated spleens following cranial radiation exposure (Koturbash et al. 2008). Th ese authors found male mice to be more susceptible to bystander eff ects in comparison to females.
Another very important phenomenon associated with non-targeted radiation eff ects is induced radioresistance responses. Th is cell protective eff ect is not unique to radiation alone, rather it has been observed with acute hypoxiainduced stimuli within analogous systems and many diff erent cell types (Michiels 2004). Th e induction of radioresistance responses were found in areas of high natural background radiation in Ramsar compared to control populations from regions of low background radiation (Mohammadi et al. 2006). Lymphocytes were extracted and exposed to 4 Gy of gamma radiation, and individuals residing in high natural radiation background areas had signifi cantly higher DNA damage and repair than control groups (Mohammadi et al. 2006). Other investigators assessing fractionated X-ray treatments, found enhanced clonogenic survival following subsequent treatments in radiosensitive clones of human colorectal tumor cell lines (Qutob et al. 2006).
In the present study, non-targeted radiation eff ects were assessed with an in vivo-based assay for blood, urine, and esophageal biopsy samples taken before and after a fractionated brachytherapy regime. Th e primary motivation of this study was to explore radiation-induced bystander eff ects (RIBE) in blood, urine, and biopsy samples taken from esophageal cancer patients undergoing fractionated HDR-ILBT. Secondary objectives were to assess whether blood and urine samples pre-exposed to one treatment fraction of brachytherapy induces radioresistance, by stimulating an increase in reporter cells survival, during subsequent exposure to brachytherapy. Additionally, certain patient characteristics were assessed to determine whether these variables are infl uencing cell communicating signals that ultimately aff ects cell cloning capabilities.
Based upon previous in vitro studies (Boyden andRaaphorst 1999, Maguire et al. 2007), it is hypothesized that fractionated treatments will induce a cell communicating protective response in reporter cells exposed to patient samples taken following each fraction of brachytherapy. Th is work will contribute to the limited data available and further our understanding of non-targeted radiation eff ects in brachytherapy at therapeutic doses.

Sample design
Blood, urine and biopsy samples were obtained from patients diagnosed with either esophageal adenocarcinoma (EA) or squamous cell carcinoma (SCC) undergoing HDR-ILBT between March 2011 and February 2012. Th e majority of patients were males diagnosed with esophageal adenocarcinoma (EA). Roughly 54% of the patients had stage III cancer and 26.7% had stage IV cancer, refer to Table I for patient characteristics and demographics. Th is research was carried out according to the Declaration of Helsinki with informed consent obtained from all participants, and ethics approval was obtained from the Hamilton Health Sciences Faculty of Health Sciences (HHS/FHS) research ethics board . In the present study, 24 patients were eligible for recruitment, however, only 11 men and four women, with a mean age of 69 years (age range, 57 -90 years) participated in the study. Out of the 15 patients, two patients discontinued from the study after the fi rst fraction of HDR-ILBT, one patient refused to undergo fraction 2 and 3 of brachytherapy, and one patient was deemed ineligible to participate after the fi rst fraction of treatment by the attending physician. All patients received 600 cGy per HDR fraction prescribed 1 cm from the source axis to the esophageal planning volume with a remote afterloading HDR unit (Varisource HDR, Nucletron, Varian International, USA) administrating high doses of gamma radiation, by using a Ir 192 source. Th e length of the treatment fi eld is determined at the time of endoscopy which occurred right before the catheter is set in place on the day of brachytherapy. Appropriate margins were set based on clinical visual determination of the tumor where a 2 cm treatment margin was added proximal and distal to the tumor. Th e dose-rate ranged anywhere between 33.1 and 109.0 Gy/h. Further details on eligibility criterion has been described in a small pilot study published elsewhere (Pinho et al. 2012).

Sample collection
Blood and urine samples were collected at the start and end of each fraction of HDR-ILBT. Tissue specimens were biopsied from the tumor-free mucosa layer of the esophagus, proximal to the tumor site. A biopsy puncture technique was used to extract tissue specimens ranging in size from 1 -2 mm 2 . Biopsies of the esophagus were obtained prior to the fi rst fraction of HDR-ILBT (baseline sample) and immediately following the fi nal fraction of treatment (test sample). Urine samples were placed in a 70 ml sterile container (Sarstedt, Montreal, QC, Canada) and peripheral blood samples in a 10 cc red lid serum Vacutainer containing no additive (BD Vacutainers, Fisher Scientifi c, Ottawa, ON, Canada). Th e samples were placed in a collection holder on ice immediately following extraction to maintain the integrity of the sample. For serum extraction, blood samples were centrifuged at 2000 rpm for 10 min and the serum was aliquoted into 5 ml sterile polypropylene tubes (Sarstedt, Montreal, QC, Canada). Th e serum was extracted from blood samples within 2 h of being collected from patients in the clinical trial laboratory at the Juravinski Cancer Centre (JCC). Biopsy samples were collected and transported in 15 ml sterile polypropylene tubes containing RPMI medium with a fi nal concentration of 200 U/ml penicillin and 200 μ g/ml streptomycin solution, 15 mM HEPES buff er, 1 μ g/ml of Fungizone, 50 μ g/ml of Nystatin, 0.5 μ g/ml of hydrocortisone, and 2 mM of L-glutamine solution. Culture medium and supplements were obtained from Invitrogen Burlington Ontario. All samples were transported on ice to our research laboratory at McMaster University and were processed within 8 h of being collected.

Cell line
Human keratinocytes HPV-G cultures (Pirisi et al. 1988) were used as a reporter to determine whether bystander signals were being generated, following HDR-ILBT, in blood serum, urine and esophageal samples. Th e human keratinocyte reporter model has been widely accepted in a number of labs to have a well-characterized and stable bystander response over a large range of doses , Mothersill et al. 2001, Ryan et al. 2008, Ahmad et al. 2013. Th e complete growth medium used for routine maintenance and colonyforming assays was RPMI-1640 with 10% foetal bovine serum (FBS) (Invitrogen, Burlington, ON, Canada), 100 U/ml penicillin and 100 μ g/ml streptomycin (Gibco, Burlington, ON, Canada), 2 mM L-Glutamine (Gibco, Burlington, ON, Canada), 0.5 μ g/ml of hydrocortisone (Sigma-Aldrich, Oakville, ON, Canada), and 15 mM of Hepes. All experiments were performed in a class II biosafety cabinet at McMaster University. Routine subculturing was performed on cell stocks reaching 80 -100% confl uency by using a 1:1 solution of 0.25% trypsin and 1 mM EDTA at 37 ° C for 8 min. Cell stocks were grown in 75 cm 2 fl asks (T-75) fi lled with 30 ml of supplemented growth medium. Cell stocks and colonyforming experiments were incubated at 37 ° C and 5% carbon dioxide in air.

Tissue explants
Prior to fraction 1 and immediately following fraction 3, biopsies were taken as described above. Tissue dissections were not needed since three biopsies were taken at fraction 1 and 3. Each piece of tissue collected was approximately 1 -2 mm 3 in size and these samples were aseptically plated in the center of 25 cm 2 fl asks (T-25) fi lled with 4 ml of supplemented growth medium. Th e complete growth medium for clonogenic assays was similar to the tissue sample collection medium except for the exclusion of antimycotics and a fi nal concentration of 100 U/ml penicillin and 100 μ g/ml streptomycin solution (Gibco, Burlington, ON, Canada) was used. Th e esophageal explants were placed in the incubator at 37 ° C in 5% carbon dioxide in air for 48 h.

Clonogenic assay Explant conditioned medium
For esophageal explants, HPV-G reporters were set up at a density of 500 cells per T-25 fl ask containing 4 ml of culture medium. Explant conditioned medium (ECM) was generated by incubating the esophageal explants in culture medium for 48 h as described above. After 48 h, a standard medium transfer was performed where the ECM was fi ltered with a 0.22 μ m Nalgene fi lter (VWR Burlington, Ontario, Canada) and placed onto reporter cultures. Following medium transfers, reporters were grown in an incubator at 37 ° C with 5% carbon dioxide in air for 10 -14 days. Once viable colonies had formed, the cells were stained with 20% carbol fuchsin (VWR, Burlington, Ontario, Canada) and colonies with Ն 50 cells were scored. Biopsies taken at the start of fraction 1 prior to the patient undergoing HDR-ILBT were used as controls. Biopsies taken immediately following irradiation were treatment samples. Seymour and Mothersill (2006) developed a blood serum in vivo colony-forming assay to assess human subject responses to radiation treatment. In the present with a linear regression analysis. A complete analysis on all 15 patients was not feasible for a number of reasons including patients leaving from the study, patients unable to give urine samples, and logistical diffi culties associated with sample collection. All p -values less than 0.05 were considered statistically signifi cant.

Results
Blood based colony-forming assay Figure 1 shows the relationship between cloning efficiency and dose-rate administered at each fraction of brachytherapy. Patients undergoing fraction 2 and 3 of brachytherapy demonstrated a signifi cant moderate positive relationship between cloning effi ciency and dose-rate ( p Ͻ 0.05 * ), whereas, fraction 1 had no such relationship. Th e association between cloning effi ciency and dose-rate were assessed further with a linear regression analysis for fractions 2 and 3 of brachytherapy. A positive trend between cloning effi ciency and dose-rate was observed for fractions 2 ( p Ͻ 0.05 * ) and fractions 3 ( p Ͻ 0.001 * * ). Th is model indicates that 29.7% and 36.3% of the total variation with the cloning ability of non-irradiated keratinocytes can be explained by the dose-rate for fractions 2 and 3, respectively ( Figure 1). Th e data is showing bystander reporters exposed to blood sera taken from cancer patients under-study, this technique was utilized on esophageal cancer patients undergoing fractionated brachytherapy. On the evening prior to patient treatments, human keratinocyte reporters were seeded at 300 cells per T-25 fl ask containing 5 ml of RPMI containing a fi nal concentration of 100 U/ml penicillin and 100 μ g/ml streptomycin (Gibco, Burlington, ON, Canada), 2 mM L-Glutamine (Gibco, Burlington, ON, Canada), 0.5 μ g/ml of hydrocortisone (Sigma-Aldrich, Oakville, ON, Canada), and 15 mM of Hepes. Th e FBS typically used in growth medium was substituted with 10% blood serum (0.5 ml per 5 ml of culture medium) collected before and after treatments. Treatments were performed early morning and blood serum was extracted at the JCC. All samples were then transported to McMaster University. Th e serum was added to the medium and then transferred to reporter fl asks. Similar to explants clonogenics, reporters were incubated at 37 ° C with 5% CO 2 in air for 10 -14 days, and then stained and colonies with Ն 50 cells were scored.

Urine samples
Th e in vivo colony-forming assay with urine samples was developed by Pinho et al. (2012) and preliminary data were published in Pinho et al. (2012). Th ese clonogenic assays were performed alongside biopsy and blood sample experiments. Human keratinocyte reporters were seeded with 700 cells in T-25 fl asks in 5 ml of RPMI culture medium. Urine samples were diluted 10-fold and added to the fl asks. 1 ml of diluted pre-and post-treatment urine samples were added to fl asks. Control fl asks with an additional 1 ml volume of diluted medium (diluted with sterile distilled water) was set up to ensure 1 ml volume of diluted urine with medium did not aff ect the colony-forming ability of reporters. Reporter cells were incubated at 37 ° C with 5% CO 2 in air for 10 -14 days, and then stained and colonies with Ն 50 cells were scored.

Statistical analysis
All reporter fl asks were set up in triplicate for each sample and at every fraction of brachytherapy. Data presented in this paper display three measurements per patient at each treatment fraction. A Shapiro-Wilk normality test and Levene ' s tests found that the data violated the normality and equal variances conditions required for a parametric statistical analysis. When assessing whether blood and urine samples repeatedly taken from patients at various time-points throughout brachytherapy had a distinct treatment eff ect, an non-parametric Friedman ' s test with a post hoc Wilcoxon signed ranks test were performed. Th e p -values were adjusted with Bonferroni corrections to eliminate the chance of committing type I errors. When before and after treatment groups were compared for patient characteristics or tissue explant clonogenic assays, signifi cance was determined by performing separate Wilcoxon signed rank tests. For the urine based colony-forming assay, there was a limited number of female patients able to give a sample. As a result, patients ' cancer staging characteristics were analyzed only. Th e relationship between cell survival and dose-rate was assessed using a Spearman ' s correlation on the blood and urine samples. Statistically signifi cant correlations were analyzed further  Figure 1. A statistically signifi cant positive relationship between cloning effi ciency (%) for HPV-G reporters and dose-rate was observed following fraction 2 and 3 of brachytherapy for blood sera samples taken from 11 patients. Outlined above are three measurements set-up per patient for each fraction of brachytherapy (n ϭ 33). For treatment fractions illustrating a signifi cant relationship between cloning effi ciency and dose-rate, a linear regression model was used to determine whether the dose-rate variable contributed to the prediction of cloning capabilities of HPV-G reporters. Fraction 2 and 3 show a clear inverse dose-rate eff ect for HPV-G reporters exposed to blood sera taken following brachytherapy. * Indicates a p -value less than 0.05.
going high dose-rate brachytherapy had a clear inverse dose-rate eff ect during late treatment fractions. Th e doserate variability observed across each of the fractions of brachytherapy can be explained in part by the patients ' tumor size, but most likely the dose-rate diff erences between fractions is related closely with the decay parameters (i.e., source decay and source renewal), refer to Supplementary data in Figure 1, available online at http://informa healthcare.com/abs/doi/10.3109/09553002.2015.1068458. When running the repeated measures analyses, brachytherapy revealed a statistically signifi cant diff erence in cloning effi ciency for HPV-G reporters treated with blood sera amongst the treatment fractions ( p Ͻ 0.001 * ), refer to Figure 2. Although when examining each treatment fraction, it can be seen that samples taken before brachytherapy compared to post-treatment samples revealed no statistical changes in cloning effi ciencies ( p Ͼ 0.05). Rather a signifi cant increase in the colony-forming ability of non-irradiated reporters was observed at the later part of brachytherapy. For instance, the fi nal fraction of brachytherapy had a statistically signifi cant increase in cloning effi ciency by 12.60% relative to baseline samples ( p Ͻ 0.001 * ). Whereas post-treatment samples at fractions 1 and 2 had an insignifi cant increase in cloning effi ciency by 7.22% ( p ϭ 0.705) and 8.65% ( p ϭ 0.210) compared to baseline samples, respectively. To eliminate the chance of committing type I errors with Wilcoxon multiple pairwise comparisons when assessing treatment eff ects at various points in time, each p -value was adjusted with Bonferroni corrections refer to (Supplementary data in Table 1, available online at http://informahealthcare.com/abs/doi/ 10.3109/09553002.2015.1068458).
When exploring the outcome of cloning effi ciency in respect to gender for samples taken before and after treatment, these patient characteristics were found to have no signifi cant infl uence on the growth of non-irradiated cells ( p Ͼ 0.05), refer to Figure 3a. In contrast to gender diff erences, cancer staging showed a signifi cant increase in cloning effi ciencies for patients clinically diagnosed with stage III at fraction 1 of treatment and stage IV at fraction 2 of brachytherapy, as shown in Figure 3b.

Urine based colony-forming assay
Th e relationship between cell survival and dose-rate was also assessed for urine samples. For patients undergoing fraction 1, 2, and 3 of brachytherapy, there was no relationship between cell survival (%) and dose-rate (Figure 4).
Similar to the blood sample data, urine samples had a statistical diff erence in cell survival throughout the course of were also assessed across each treatment group to explore whether any changes in cloning effi ciency (%) were dependent on certain patient characteristics. For each fraction, separate Wilcoxon signed rank tests were performed to determine whether there are signifi cant changes between males and females and cancer stage III and IV. Since 3 measurements were set-up per patient, n ϭ 33 for males, n ϭ 12 for females, n ϭ 24 for patients diagnosed with cancer stage III and n ϭ 12 for cancer stage IV. All values are mean Ϯ SEM. * Indicates statistically signifi cant diff erence from pre-treatment sample.
Instead of the eff ect occurring in the post-treatment samples at the fi nal fraction of treatment, samples taken before the third fraction of brachytherapy had a signifi cant increase in cell survival compared to post-treatment samples after the fi rst fraction of brachytherapy ( p Ͻ 0.001 * ). In Figure 5, it can also be seen that samples taken after fractions 1 and 2 had insignifi cant changes in cell survival compared to baseline samples. Similar to the blood sample results, these fi ndings are suggesting that late treatment fractions of brachytherapy are inducing a radioresistance response in non-irradiated cells. Furthermore, cancer staging had no infl uence on the growth of non-irradiated cells ( p Ͼ 0.05), however, the data reveals large variability for this endpoint ( Figure 6). brachytherapy ( p Ͻ 0.05 * ), as shown in Figure 5 and Supplementary data in Table 2. available online at http://informa healthcare.com/abs/doi/10.3109/09553002.2015.1068458 A signifi cant increase in the survival of non-irradiated reporters was also observed at the later part of brachytherapy. . Shown above is the HPV-G reporters ' clonogenic survival (%) after being exposed to explant conditioned medium with samples taken before the fi rst fraction and immediately after the fi nal fraction of brachytherapy. Six patients had triplicate fl ask set-up for each treatment. All values are mean Ϯ SEM of n ϭ 18.
Side-eff ects in radiotherapy regimes are primarily attributed to diff erent patients having inherently unique radiosensitivities (Twardella and Chang-Claude 2002). One of the fi rst promising studies assessing the RIBE and cancer patients ' intrinsic radiosensitivities from blood samples was published by Howe et al. (2009). In this study, it was shown that lymphocyte cultures, taken from colorectal cancer patients, had a signifi cant increase in radiosensitivity and its ability to produce bystander signals compared to cancer-free controls. Other studies detected bystander and radioprotective factors in the blood serum of Chernobyl accident survivors (Marozik et al. 2007) and cancer patients undergoing various fractions of radiotherapy , respectively.
Bystander signalling has been suggested to be associated with the activation of macrophages in mice . Recently researchers have shown that radiation stimulates the innate immune function , Multhoff and Radons 2012, R ö del et al. 2012, Mothersill 2013. With such high doses being prescribed at each fraction, further investigation with a macrophage Superoxide Dismutase (SOD) assay revealing innate immune function may be benefi cial to test in future work (Johnston et al. 1978, Fukasawa et al. 1988. Th e superoxide anion (0 2 -) is a short-lived free radical that plays an essential role in immune responses (Johnston et al. 1978). Such a radical is commonly released from macrophages. Macrophages collected from blood and cultured using regular cell culture techniques would be one way of investigating whether this response is a systematic immune response. A more mechanistic approach to elucidate other cellular activities would be to assess reactive oxygen species (ROS) activity in the bystander cells by using the 2 ′ ,7 ′ -Dichlorofl uorescein (DCF) fl uorescent probe (O ' Dowd et al. 2006). Th is marker for ROS can be loaded into the human keratinocytes cells after exposure to medium supplemented with 10% blood serum taken from cancer patients undergoing brachytherapy. Past work has found that increased levels of fl uorescence has been highly correlated with higher levels of ROS in bystander cells (O ' Dowd et al. 2006).
From our experiments, the urine-based colony-forming assays showed substantial inter-patient variability relative to the blood based assay results. Th e urine-based assay had proved to be unreliable and was deemed unsuitable for further clinical work due to the large variation observed amongst patients and treatment fractions. However, the blood based assay had much less variability and revealed interesting fi ndings that provides further insight on the previously published work Mothersill 2006, Marozik et al. 2007). Unlike previous in vitro work resulting in a lower ability to produce bystander signal(s) when tissue samples were harvested from males with a pre-existing malignancy (Mothersill et al. 2002), the present clinical study had no such infl uence on signal production for gender. Furthermore cancer staging had no observable infl uence on the growth of non-irradiated reporters in cancer patients undergoing brachytherapy for the urine samples, but there was a signifi cant eff ect observed for cancer staging for blood serum data only. Th ese inconsistent results are most likely attributed to the small sample size and would need to be assessed further with a larger sample.

Tissue explant based colony-forming assay
For the biopsy samples, tissues taken at baseline had a lower survival by 6.00% compared to samples taken immediately following the fi nal treatment of brachytherapy ( p Ͼ 0.05). Th ese results reveal no indication of tissue explants generating bystander signals following brachytherapy.

Discussion
Th e primary objective of this study is to determine whether blood, urine, and tissue explant-based colony forming assays can be used to trace levels of bystander or protective signals being generated following brachytherapy treatments. A few patient characteristics were assessed to determine whether these factors might be infl uencing cell communicating signals and aff ecting the growth of non-irradiated cells.
Although the data is limited, the fi ndings may suggest that cancer patients undergoing fractionated brachytherapy induced a radioresistance response for cells or tissues in close proximity to the irradiated tumor volume after undergoing treatment 3 compared to baseline samples. Such a response was observed in bystander cells exposed to blood sera and urine samples taken from esophageal cancer patients irradiated in vivo. Urine sample results had a similar trend as the blood serum data, although, these samples did not have a signifi cant increase in cell survival when taken immediately after fraction 3. Instead, urine samples taken before fraction 3 had a signifi cant increase in the reporters cell survival compared to post-treatment samples taken after the fi rst fraction of brachytherapy.
In the literature, there is a considerable amount of research on cellular radioresistance responses performed in vitro (Th omas et al. 2013) , once cells have been exposed to small acute doses or low acute dose-rates, initiating protective responses or enhanced repair processes. However, the induced radioresistance response is commonly triggered with doses below 1 Gy and dose-rates ranging from 0.18 -2.43 Gy/min (Th omas et al. 2013). In the present study, irradiations occurred in vivo with a prescribed dose to the esophageal lumen of 600 cGy per HDR fraction and the average dose-rates per fraction were Ͼ 50 Gy/h. Our fi ndings are suggesting an induced radioresistance response after subsequent treatment fractions. However, a characteristic of this phenomenon is that low doses below 1 Gy are required during in vitro irradiations. One possible explanation for the eff ect being triggered late into brachytherapy regimes with substantially higher doses and dose-rates, would be that circulating blood cells fl owing through the tumor volume may have been directly irradiated with signifi cantly lower doses of gamma radiation than the tumor itself. However, there is also a possibility that blood cells fl owing nearby the tumor spend signifi cantly less time in the radiation fi eld and may have not been directly irradiated, rather the eff ect may be a systemic immune response (Mothersill and Seymour 2004). Th e response observed in the non-irradiated cells, incubated with supernatants of blood serum during the fi nal fraction of brachytherapy, are presumably initiated as a result of neighboring cells receiving comparable doses to in vitro radiation studies. a predictive assay for assessing radiation side-eff ects or treatment outcome. A follow-up study has been undertaken with a target sample size of 115 cancer patients and 15 healthy patients with a power (1-β ) set at 0.95 and statistical significance level set to 0.05. Th is work will provide further insight on whether non-targeted radiation eff ects have relevance in HDR brachytherapy.
Th e data also revealed a clear statistically signifi cant inverse dose-rate response in bystander cells exposed to blood serum harvested immediately following brachytherapy at fractions 2 and 3. However, the urine samples had no such eff ect observed following fraction 2 and 3 treatments. An earlier study by Mitchell et al. (1979), observed an inverse doserate eff ect during in vitro radiations of HeLa cells exposed to dose-rates ranging from 0.37 -1.54 Gy/h (Mitchell et al. 1979). Th ese authors found that HeLa cells had an increase in cell death for lower dose-rates than higher ones. At certain doserates, HeLa cells progress through the cell cycle and become blocked in the radiosensitive G2 phase at lower dose-rates, resulting in enhanced cell killing.
One of the most problematic issues that should be addressed in this paper were the loss of valuable biopsy samples, due to patient contamination. Patient contaminations were attributed to opportunistic yeast infections such as Candida albicans , which are common in immunocompromised cancer patients as explained in Delsing et al. (2012). Due to the scarcity of data available on such infections for esophageal carcinoma patients (Chiou et al. 2000), these issues were originally overlooked at the start of the clinical study. Th roughout the remainder of the sample collection, antimycotics (fungizone and nystatin) were supplemented in the collection medium prior to commencing the tissue explant colony-forming assays as a preventive measure. It is our recommendation for other investigators to either check for yeast infections or to supplement fungizone and nystatin for a short duration of time to avoid similar problems. As it stands, the tissue explant colony-forming assays led to inconclusive data and even more unanswered questions. Further investigations are warranted to assess the radiosensitivity and non-targeted radiations eff ects in nearby esophagus samples.
Another limitation associated with this study was the small sample size and the substantial inter-patient variation observed from one treatment fraction to the next. Th e limited sample size was attributed to diffi culties with consenting patients for biopsy samples, as well as for all stages of treatment, and 3 weeks follow-ups (unreported data). Th e number of subjects required to achieve statistical power was conducted with a power analysis using G * Power software (Faul et al. 2007) with a power (1-β ) set at 0.95 and α ϭ 0.05. Future work would need to have a sample size of 84 in order to determine whether the diff erent treatment fractions changes in cell survival have reached statistical signifi cance at the 0.05 level. In retrospect, a much simpler study design focusing on only one treatment fraction and blood serum samples alone would be more appropriate since esophageal cancer typically presents at advance stages where the diseases prognosis are quite poor (Sur et al. 1998).
In conclusion, this simple blood-based assay may have future implications as a clinical detection tool used to predict treatment outcome based upon certain patient characteristics, such as gender, cancer staging, and metastatic status. From these preliminary fi ndings, a follow-up study with a larger sample size, including unirradiated cancer-free controls, and a thorough analysis of patient characteristics may shed light on whether this technique may be appropriate as