- Email: [email protected]
Differential gene expression before and after ionizing radiation of subcutaneous fibroblasts identifies breast cancer patients resistant to radiation-induced fibrosis
Radiotherapy and Oncology 83 (2007) 261–266 www.thegreenjournal.com
Clinical radiobiology
Differential gene expression before and after ionizing radiation of subcutaneous fibroblasts identifies breast cancer patients resistant to radiation-induced fibrosis Jan Alsnera,*, Olaug K. Rødningenb, Jens Overgaarda a
Department of Experimental Clinical Oncology, Aarhus University Hospital, Denmark, bDepartment of Genetics, University of Oslo, Norway
Abstract Background and purpose: Differentially gene expression between patients with either very low or very high risk of radiation-induced fibrosis (RIF) in patient-derived fibroblasts after irradiation has previously been reported. In the present study, we are investigating the robustness of radiation-induced changes in gene expression in fibroblasts, whether differential expression is more pronounced when looking at the fold induction levels, taking into account the differences in background expression levels between patients, and whether there is a linear correlation between individual risk of RIF and changes in radiation-induced gene expression in fibroblasts. Material and methods: Gene expression was analysed by quantitative real-time PCR before and after a fractionated scheme with 3 · 3.5 Gy/3 days in fibroblasts derived from 26 patients with breast cancer treated with post-mastectomy radiotherapy. Results: Robust radiation-induced changes in gene expression were observed, with differential gene expression between low and high risk patients being most pronounced for the fold induction level (‘after’ value divided by ‘before’ value for each patient). When including patients with intermediate risk, there was no linear correlation between individual risk of RIF and differential expression of the genes investigated. Rather, differential gene expression could divide patients into two clearly separated groups, a larger, sensitive group and a smaller resistant group. Conclusions: Differential gene expression in irradiated fibroblasts might be an important tool in the identification of differences in the genetic background between patients with variable risk of RIF, and in the identification of new targets for prevention and intervention of the fibrotic process. c 2007 Elsevier Ireland Ltd. All rights reserved. Radiotherapy and Oncology 83 (2007) 261–266.
Keywords: Ionizing radiation; Radiation-induced fibrosis; Differential gene expression; Fibroblasts; Prediction
Radiation-induced fibrosis (RIF) is a frequent late side effect of radiation therapy, and can be a major dose-limiting factor [1–3]. Like other normal tissue complications after radiotherapy, the risk of RIF shows a large degree of variability between patients [4], and is known to be affected by differences in treatment characteristics [5–7] as well as a number of patient related factors, like age, nutritional status, medication, coexisting morbidity, and recent surgery [4]. In addition to these factors, it seems likely that clinical variability in radiation sensitivity can be attributed to profound biological differences between patients [8]. It has long been hoped that identification of these biological differences will lead to more individualized therapeutic strategies and greater understanding of the pathology. Previous attempts to develop predictive assays for RIF have mainly focused on in vitro radiosensitivity and DNA damage endpoints. Although some reports have been promising [9,10], it is still an open question whether assays based
solely on cell killing and DNA damage can be developed into clinically useful predictors [11]. Furthermore, these studies do not provide any new knowledge about the biology behind RIF [11]. More recently, a number of reports have shown that clinically relevant variability in RIF may be associated with different genetic variants [12]. These studies have focused on relatively few single nucleotide polymorphisms (SNPs) in a limited number of candidate genes identified from existing knowledge about the molecular pathology of RIF [13], but have provided important evidence to support the idea that genetic variability is important for the risk of RIF in unselected cancer patients [12]. Microarray-based gene expression studies on the presumed target cell, fibroblasts, can provide new knowledge about genes and pathways affected by various types and doses of irradiation [14–19]. Interestingly, one study comparing ionizing radiation from an external source with DNA-incorporated 125I found a 10-fold higher number of
0167-8140/$ - see front matter c 2007 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.radonc.2007.05.001
262
Fibroblast gene expression and fibrosis
genes induced by external c-irradiation, although the amount of DNA double-strand breaks was higher in cells with DNA-incorporated 125I [20]. Using cDNA microarray data from patient-derived fibroblasts before and after in vitro irradiation, we have recently been able to identify genes, unrelated to DNA damage response, which are differentially expressed between patients with either low or high risk of RIF in patient-derived fibroblasts after irradiation [21]. Differential expression was only observed in the irradiated fibroblasts and only when we looked at two groups of patients, very low risk patients compared to very high risk patients. Interestingly, genes that were generally induced by irradiation were all less upregulated – and genes generally repressed were all less downregulated – in the low risk patients. In the present study, the robustness of radiation-induced changes in gene expression in fibroblasts is addressed by using new batches of cells, new irradiations, different RNA extraction and cDNA synthesis procedures, and measuring gene expression by a different technology (real-time PCR). Furthermore, patients with intermediate to high risk are included, in order to test whether there is a linear correlation between the radiation-induced changes in gene expression and the individual risk of RIF. Finally, differential expression patterns are analysed at the fold induction levels, taking into account the differences in background expression levels between patients.
Cell culture, irradiation, RNA extraction and reverse transcription Fibroblasts were expanded and irradiated as described previously with the exception that only the fractionated scheme of 3 · 3.5 Gy in intervals of 24 h was used [21]. RNA was extracted 2 h after the last fractionated dose using the RNeasy Mini Kit (Qiagen) according to the manufacturers details. RNA was quantified using a spectrophotometer (GeneQuant, Pharmacia Biotech) and 2 lg total RNA was used to generate cDNA with the High Capacity cDNA Archive kit (Applied Biosystem) according to the manufacturers instructions.
Quantitative real-time PCR TaqMan gene expression assays (Applied Biosystems) were used to quantify the following transcripts (gene symbol and Assay ID): ARID5B, Hs01381961_m1; CDC6, Hs00154374_m1; CDON, Hs00610901_m1; CTGF, Hs00170014_m1; CXCL12, Hs00171022_m1; DEGS1, Hs00186447_m1; FAP, Hs00189476 _m1; FBLN2, Hs01002063_m1; LMNB2, Hs00383326_m1; LUM, Hs00158940_m1; MT1X, Hs00745167_sH; MXRA5, Hs00377849_m1; PLAGL1, Hs00243030_m1; PMM1, Hs00160 195_m1; SLC1A3, Hs00188193_m1; SOD2, Hs00167309_m1; SOD3, Hs00162090_m1; and WISP2, Hs00180242_m1. For each reaction, cDNA (corresponding to 5 ng total RNA), 1· assay mix and 1· TaqMan Universal PCR Mastermix (AppliedBiosystems) were mixed in a total of 25 ll. All reactions were performed in duplicate on an ABI Prism 7000 Sequence Detector (AppliedBiosystems).
Materials and methods Patient material
Data analysis
The ‘Aarhus post-mastectomy cohort’ consists of 319 breast cancer patients treated with post-mastectomy radiotherapy between 1978 and 1982 [22,23]. The 26 patients included in the present study are from a smaller cohort of 41 patients from whom cultured fibroblasts are available [24,25]. The patients were treated with an anterior photon field to the axillary and infra/supraclavicular regions, a wax bolus covering the surgical scar in the axilla and with an anterior electron field to the chest wall. In every field, the physical dose absorbed at a dosimetric reference depth of 4.1 mm was calculated and radiation doses were converted into the equivalent biological dose of 2 Gy per fraction using a linear-quadratic model [24]. The evaluation of subcutaneous fibrosis was done by superficial palpation in each of the three treatment fields, and the severity of fibrosis was graded on a four point ordinal scale (grades 0–3) [22] identical to that later used in the LENT-SOMA (Late Effects on Normal Tissues – Subjective, Objective, Management and Analytic) scoring system. Excess risk of fibrosis (an estimate of how the response in a patient compares with the average response in the cohort of 41 patients) is defined as the difference between the observed response in each treatment field of the individual patient and the expected response and was calculated in each field from a dose–response curve based on the 41 patients, taking into account the variation in total dose, dose per fraction and followup time for each patient. The average over the three treatment fields was used to obtain a single measure of clinical radiosensitivity [25,26].
Gene expression levels were normalized to PMM1, phosphomannomutase 1, which based on cDNA microarray data is not affected by irradiation [21]. All assays were tested against PMM1 for equal amplification efficiency based on cDNA standard curves (generated from a mixture of cDNAs from various treated and untreated samples). Twelve assays could be analysed using the comparative CT method (CDON, CTGF, DEGS1, FBLN2, LMNB2, LUM, MT1X, MXRA5, PLAGL1, SLC1A3, SOD2, SOD3) whereas the remaining five were analysed using cDNA standard curves. Fold induction levels were calculated by dividing the relative expression of a given gene to PMM1 in the irradiated sample by the relative expression of the same gene to PMM1 in the unirradiated sample. Two-way clustering was performed using complete linkage hierarchical clustering (rana.lbl.gov/EisenSoftware.htm) of geometric mean centred data, and displayed with Java TreeView (jtreeview.sourceforge.net/). Based on corresponding observations of equivalent biological dose and scoring of normal tissue reaction in each radiation field, dose–response curves were established for subcutaneous fibrosis as described previously [24]. Briefly, the observations were grouped into eight groups according to the equivalent biological dose absorbed at the reference depth. Due to the lower number of observations, the observations were grouped into five and four groups when patients were separated into sensitive and resistant groups, respectively. The incidence of either grade 2 (moderate) or 3 (severe) subcutaneous fibrosis was plotted against the mean radiation dose in each of the dose groups.
J. Alsner et al. / Radiotherapy and Oncology 83 (2007) 261–266
Dose–response curves were fitted to the observations using logit analysis. ED50 values (defined as the radiation dose which would on the average be expected to cause a certain normal tissue effect in 50% of the treatment fields) were established for each curve. Differences in radiosensitivity between genotypes were quantified in terms of enhancement ratios at ED50 level with 95% confidence intervals.
Results In previous studies, cDNA microarray analysis was used to evaluate gene expression in irradiated subcutaneous fibroblasts from breast cancer patients treated with adjuvant radiotherapy [19,21]. By SAM (Significance Analysis of Microarrays), 60 genes were identified that were differentially expressed at >2-fold level between two groups of patients with very low and very high risks of RIF, respectively [21]. Also, by PAM (Prediction Analysis of Microarrays), a minimum set of 18 genes was identified that could differentiate between the two extreme groups [21]. In the present study, we initially selected 17 of these genes for further analysis. Ten genes (ARID5B, CDC6, CDON, CXCL12, DEGS1, LUM, MT1X, MXRA5, PLAGL1 and SLC1A3) were from the PAM gene set, and seven genes (CTGF, FAP, FBLN2, LMNB2, SOD2, SOD3 and WISP2) were selected from the genes identified by SAM. First, new batches of fibroblasts from the same previously studied four patients with very low risk and 10 patients with very high risk of RIF [21] were expanded and irradiated according to the same procedures [19]. Gene expression was measured by real-time quantitative PCR using PMM1 as endogenous control [21] before and after irradiation. In accordance with the cDNA array data [19], 12 of the genes were upregulated by fractionated irradiation (ARID5B, CDON, CXCL12, FAP, FBLN2, LUM, MXRA5, PLAGL1, SLC1A3, SOD2, SOD3 and WISP2) and five genes were downregulated (CDC6, CTGF, DEGS1, LMNB2 and MT1X) (data not shown). Also in accordance with the cDNA array data [21], the 12 genes that were induced by irradiation were less upregulated in the group of fibroblasts from very low risk patients compared to the group of very high risk patients, and likewise, the five repressed genes were less downregulated in fibroblasts from very low risk patients (data not shown). By dividing the relative values of each gene to the endogenous control gene (PMM1) in the irradiated samples with the corresponding relative value in the unirradiated samples, a fold induction level was calculated for all 17 genes. In order to select genes with high differential expression levels, the geometric mean of the fold induction levels in the group of very low risk patients versus the group of high risk patients was compared. For 13 genes (CDC6, CDON, CXCL12, FAP, FBLN2, LMNB2, LUM, MT1X, MXRA5, SLC1A3, SOD2, SOD3 and WISP2) the difference in induction levels was more than 2-fold (data not shown), and these genes were selected for further analysis. The 14 very low and very high risk patients are part of a cohort of 41 patients where fibroblasts as well as data on excess risk of RIF are available (see Materials and methods for details). Next, we extended and irradiated fibroblasts
263
from an additional 12 patients with intermediate to high excess risk scores, and measured gene expression levels for the 13 genes. Fig. 1 illustrates the expression levels before and after irradiation, as well as the fold induction levels, of four representative genes that are either upregulated (MXRA5 (Fig. 1a), and FAP (Fig. 1b)) or downregulated (MT1X (Fig. 1c) and LMNB2 (Fig. 1d)) in fibroblasts from these 26 patients. Expression levels and fold induction levels are plotted against excess risk scores with negative values indicating low risk and positive values indicating high risk of RIF. With the possible exception of MXRA5, none of the 13 genes showed any correlations between expression levels in unirradiated samples and excess risk scores. Differential expression was only observed following irradiation and was most pronounced for the fold induction levels. None of the genes showed any linear correlation between excess risk score and fold induction levels. In general, fibroblasts from patients with very low excess risk scores tended to be least affected by irradiation (having fold induction levels close to 1), whereas the majority of patients with intermediate to high excess risk scores were induced (or repressed) to higher (or lower), but similar, levels. To perform hierarchical clustering analysis of the fold induction levels (Fig. 2), the fold induction value for each gene was divided by the geometric mean value of all fold induction values for that gene. Thus, for the upper gene cluster, which include genes that are generally repressed by irradiation, high values (red) indicate samples with fold induction values closer to 1.0, i.e. less downregulated by irradiation. The lower gene cluster, which include genes that are generally induced by irradiation, low values (green) indicate samples with fold induction values closer to 1.0, i.e. less upregulated by irradiation. Two distinct clusters of patients were identified. The small cluster to the right containing seven patients include the four most resistant patients with the lowest excess risk scores. A dose–response curve was established for the risk of moderate or severe fibrosis in all 26 patients (Fig. 3, left). Patients were then divided into a sensitive and resistant group, respectively, based on the hierarchical clustering analysis in Fig. 2. From the resulting ED50 values (Fig. 3, right), an enhancement ratio of 1.25 (1.12–1.40, 95% confidence interval) was calculated.
Discussion The data presented here first of all demonstrate the robustness of radiation-induced changes in gene expression in fibroblasts. Using new batches of irradiated cells, different RNA extraction and cDNA synthesis procedures, and real-time PCR to measure gene expression, we initially attempted to reproduce expression data generated by cDNA microarrays [21]. In fibroblasts from two groups of patients with either very low or very high risk of RIF, respectively, a perfect correlation was observed between the overall effect of fractionated irradiation (induction or repression) [19]. The 17 genes analysed were selected as they were shown to be differentially expressed between the two groups with
264
Fibroblast gene expression and fibrosis
a
8.00 4.00
16.00 MXRA5
Before After
2.00 1.00 0.50 0.25 0.13 0.06
4.00 2.00 1.00
0.03 0.02 -200 -150 -100 -50
0
0.50 -200 -150 -100 -50
50 100 150 200
Excess risk score
b
4.00
2.00 1.00 0.50
2.00
1.00
0.25 0.13 0
0.50 -200 -150 -100 -50
50 100 150 200
Excess risk score
50 100 150 200
4.00 MT1X
Before After
MT1X
Difference 2.00
2.00 Fold induction
Fold difference relative to PMM1
4.00
0
Excess risk score
8.00
1.00 0.50 0.25
1.00 0.50 0.25
0.13 0.13
0.06 0.03 -200 -150 -100 -50
0
0.06 -200 -150 -100 -50
50 100 150 200
Excess risk score
16.00
0
50 100 150 200
Excess risk score 4.00
32.00 LMNB2
Before After
LMNB2
Difference 2.00
8.00 Fold induction
Fold difference relative to PMM1
FAP
Difference
4.00
0.06 -200 -150 -100 -50
d
50 100 150 200
8.00 FAP
Before After
Fold induction
Fold difference relative to PMM1
8.00
0
Excess risk score
16.00
c
MXRA5
Difference 8.00 Fold induction
Fold difference relative to PMM1
16.00
4.00 2.00 1.00
1.00 0.50 0.25
0.50 0.13
0.25 0.13 -200 -150 -100 -50
0
50 100 150 200
Excess risk score
0.06 -200 -150 -100 -50
0
50 100 150 200
Excess risk score
Fig. 1. Fibroblast gene expression levels before (s) or after 3 · 3.5 Gy/3 days (d) (left), and fold induction levels ( ) (right), for MXRA5 (a), FAP (b), MT1X (c) and LMNB2 (d) plotted against excess risk scores – see text for details.
the induced genes being less induced (and repressed genes being less repressed) in fibroblasts from the resistant group of patients [21]. All 17 genes showed a similar pattern of differential expression, although the degree of differential expression did vary slightly, with some genes showing higher
and some genes showing lower levels of differential expression. Also in accordance with the cDNA data, the expression levels before irradiation varied between patients, but did not show any differential expression between very low and very high risk patients.
J. Alsner et al. / Radiotherapy and Oncology 83 (2007) 261–266
265
Fig. 2. Hierarchical clustering analysis of the fold induction levels relative to the geometric mean values. Patients are identified by the excess risk score as in Fig. 1. 100 Moderate/severe fibrosis (%)
Moderate/severe fibrosis (%)
100 80 60 40 20 0
80 60
Enhancement ratio 1.25 (1.12-1.40)
40 20 0
30
40
50
60
70
80
Radiation dose (Gy) (equivalent dose of 2 Gy per fraction)
30
40
50
60
70
80
Radiation dose (Gy) (equivalent dose of 2 Gy per fraction)
Fig. 3. Dose–response curves for subcutaneous fibrosis in all patients ( ) or divided into groups of sensitive (d) and resistant (s) patients based on gene expression profiles. The enhancement ratio is the ratio between ED50 values (defined as the radiation dose which on average is expected to cause moderate or severe fibrosis in 50% of the treatment fields) with 95% confidence intervals.
Having established the robustness of the assay, 13 genes with the most distinct patterns of differential expression were selected for further analysis. As the previous study only concerned two extreme groups of patients, we next analysed expression patterns in 12 additional patients with intermediate to high risk of RIF. When plotting data before and after irradiation together with the fold induction level (‘after’ value divided by ‘before’ value for each patient), the most pronounced differential expression was observed for the fold induction levels (Fig. 1). It is not known whether the variations in ‘before’ expression are caused by slight differences in experimental setup (e.g. cell density) or by differences in genetic background between patients. However, the effect of fractionated irradiation does not seem to be related to background expression levels, and future studies on any differential effect of irradiation should include background expression measurements. Finally, we could not detect any linear correlation between individual risk of RIF and differential expression of the genes investigated. Rather, a cluster analysis suggests that patients can be divided in two groups with distinct expression patterns. The smaller group contains the four
most resistant patients, as defined by their excess risk score for RIF, together with a few patients with intermediate to high excess risk scores (Fig. 2). Although these latter patients are classified as more or less sensitive by their excess risk score, a dose–response curve (Fig. 3) shows that a cluster analysis based on differential gene expression is capable of dividing patients into two clearly separated groups, a larger, sensitive group and a smaller resistant group. The gene signature presented here does not necessarily represent the most optimal signature for predicting resistance to RIF. Further and larger studies are clearly needed before such a profile can be identified and tested for clinical accuracy. However, with the current experimental setup including isolation and expansion of patient-derived fibroblasts, radiation-induced changes of gene expression in vitro is not an assay that is likely soon to be included in any treatment planning phase. The potential of these studies is rather as an important tool in the identification of the differences in genetic background which leads to the differential induction and repression of genes and pathways between patients with variable risk of RIF, and in the identification of new targets for prevention and intervention of the fibrotic process.
266
Fibroblast gene expression and fibrosis
* Corresponding author. Jan Alsner, Department of Experimental Clinical Oncology, Aarhus University Hospital, Denmark. E-mail address: [email protected] Received 25 March 2007; received in revised form 1 May 2007; accepted 1 May 2007; Available online 23 May 2007
References [1] O’Sullivan B, Levin W. Late radiation-related fibrosis: pathogenesis, manifestations, and current management. Semin Radiat Oncol 2003;13:274–89. [2] Gothard L, Cornes P, Brooker S, et al. Phase II study of vitamin E and pentoxifylline in patients with late side effects of pelvic radiotherapy. Radiother Oncol 2005;75:334–41. [3] Haddad P, Kalaghchi B, mouzegar-Hashemi F. Pentoxifylline and vitamin E combination for superficial radiation-induced fibrosis: a phase II clinical trial. Radiother Oncol 2005;77:324–6. [4] Bentzen SM, Overgaard J. Patient-to-patient variability in the expression of radiation-induced normal tissue injury. Semin Radiat Oncol 1994;4:68–80. [5] Bentzen SM, Christensen JJ, Overgaard J, Overgaard M. Some methodological problems in estimating radiobiological parameters from clinical data. Alpha/beta ratios and electron RBE for cutaneous reactions in patients treated with postmastectomy radiotherapy. Acta Oncol 1988;27:105–16. [6] Kurtz JM. The clinical radiobiology of breast cancer radiotherapy. Radiother Oncol 2005;75:6–8. [7] Yarnold J, Ashton A, Bliss J, et al. Fractionation sensitivity and dose response of late adverse effects in the breast after radiotherapy for early breast cancer: long-term results of a randomised trial. Radiother Oncol 2005;75:9–17. [8] Andreassen CN, Alsner J, Overgaard J. Does variability in normal tissue reactions after radiotherapy have a genetic basis – where and how to look for it? Radiother Oncol 2002;64:131–40. [9] Borgmann K, Roper B, El-Awady R, et al. Indicators of late normal tissue response after radiotherapy for head and neck cancer: fibroblasts, lymphocytes, genetics, DNA repair, and chromosome aberrations. Radiother Oncol 2002;64:141–52. [10] Hoeller U, Borgmann K, Bonacker M, et al. Individual radiosensitivity measured with lymphocytes may be used to predict the risk of fibrosis after radiotherapy for breast cancer. Radiother Oncol 2003;69:137–44. [11] Russell NS, Begg AC. Predictive assays for normal tissue damage. Radiother Oncol 2002;64:125–9. [12] Andreassen CN. Can risk of radiotherapy-induced normal tissue complications be predicted from genetic profiles? Acta Oncol 2005;44:801–15. [13] Bentzen SM. Preventing or reducing late side effects of radiation therapy: radiobiology meets molecular pathology. Nat Rev Cancer 2006;6:702–13.
[14] Ding LH, Shingyoji M, Chen F, et al. Gene expression profiles of normal human fibroblasts after exposure to ionizing radiation: a comparative study of low and high doses. Radiat Res 2005;164:17–26. [15] Tachiiri S, Katagiri T, Tsunoda T, Oya N, Hiraoka M, Nakamura Y. Analysis of gene-expression profiles after gamma irradiation of normal human fibroblasts. Int J Radiat Oncol Biol Phys 2006;64:272–9. [16] Ishikawa K, Koyama-Saegusa K, Otsuka Y, et al. Gene expression profile changes correlating with radioresistance in human cell lines. Int J Radiat Oncol Biol Phys 2006;65:234–45. [17] Kis E, Szatmari T, Keszei M, et al. Microarray analysis of radiation response genes in primary human fibroblasts. Int J Radiat Oncol Biol Phys 2006;66:1506–14. [18] Heinloth AN, Shackelford RE, Innes CL, et al. Identification of distinct and common gene expression changes after oxidative stress and gamma and ultraviolet radiation. Mol Carcinog 2003;37:65–82. [19] Rødningen OK, Overgaard J, Alsner J, Hastie T, BørresenDale AL. Microarray analysis of the transcriptional response to single or multiple doses of ionizing radiation in human subcutaneous fibroblasts. Radiother Oncol 2005;77:231–40. [20] Sokolov MV, Smirnova NA, Camerini-Otero RD, Neumann RD, Panyutin IG. Microarray analysis of differentially expressed genes after exposure of normal human fibroblasts to ionizing radiation from an external source and from DNA-incorporated iodine-125 radionuclide. Gene 2006;382:47–56. [21] Rødningen OK, Alsner J, Hastie T, Overgaard J, Børresen-Dale AL. Application of microarray analyses to identify genes involved in radiation-induced fibrosis. Breast Cancer Res 2005;7(Suppl 2):542–3. [22] Overgaard M, Bentzen SM, Christensen JJ, Madsen EH. The value of the NSD formula in equation of acute and late radiation complications in normal tissue following 2 and 5 fractions per week in breast cancer patients treated with postmastectomy irradiation. Radiother Oncol 1987;9:1–11. [23] Bentzen SM, Overgaard M, Thames HD, Christensen JJ, Overgaard J. Early and late normal-tissue injury after postmastectomy radiotherapy alone or combined with chemotherapy. Int J Radiat Biol 1989;56:711–5. [24] Andreassen CN, Alsner J, Overgaard M, Overgaard J. Prediction of normal tissue radiosensitivity from polymorphisms in candidate genes. Radiother Oncol 2003;69:127–35. [25] Johansen J, Bentzen SM, Overgaard J, Overgaard M. Relationship between the in vitro radiosensitivity of skin fibroblasts and the expression of subcutaneous fibrosis, telangiectasia, and skin erythema after radiotherapy. Radiother Oncol 1996;40:101–9. [26] Bentzen SM, Overgaard M, Overgaard J. Clinical correlations between late normal tissue endpoints after radiotherapy: implications for predictive assays of radiosensitivity. Eur J Cancer 1993;29A:1373–6.
Clinical radiobiology
Differential gene expression before and after ionizing radiation of subcutaneous fibroblasts identifies breast cancer patients resistant to radiation-induced fibrosis Jan Alsnera,*, Olaug K. Rødningenb, Jens Overgaarda a
Department of Experimental Clinical Oncology, Aarhus University Hospital, Denmark, bDepartment of Genetics, University of Oslo, Norway
Abstract Background and purpose: Differentially gene expression between patients with either very low or very high risk of radiation-induced fibrosis (RIF) in patient-derived fibroblasts after irradiation has previously been reported. In the present study, we are investigating the robustness of radiation-induced changes in gene expression in fibroblasts, whether differential expression is more pronounced when looking at the fold induction levels, taking into account the differences in background expression levels between patients, and whether there is a linear correlation between individual risk of RIF and changes in radiation-induced gene expression in fibroblasts. Material and methods: Gene expression was analysed by quantitative real-time PCR before and after a fractionated scheme with 3 · 3.5 Gy/3 days in fibroblasts derived from 26 patients with breast cancer treated with post-mastectomy radiotherapy. Results: Robust radiation-induced changes in gene expression were observed, with differential gene expression between low and high risk patients being most pronounced for the fold induction level (‘after’ value divided by ‘before’ value for each patient). When including patients with intermediate risk, there was no linear correlation between individual risk of RIF and differential expression of the genes investigated. Rather, differential gene expression could divide patients into two clearly separated groups, a larger, sensitive group and a smaller resistant group. Conclusions: Differential gene expression in irradiated fibroblasts might be an important tool in the identification of differences in the genetic background between patients with variable risk of RIF, and in the identification of new targets for prevention and intervention of the fibrotic process. c 2007 Elsevier Ireland Ltd. All rights reserved. Radiotherapy and Oncology 83 (2007) 261–266.
Keywords: Ionizing radiation; Radiation-induced fibrosis; Differential gene expression; Fibroblasts; Prediction
Radiation-induced fibrosis (RIF) is a frequent late side effect of radiation therapy, and can be a major dose-limiting factor [1–3]. Like other normal tissue complications after radiotherapy, the risk of RIF shows a large degree of variability between patients [4], and is known to be affected by differences in treatment characteristics [5–7] as well as a number of patient related factors, like age, nutritional status, medication, coexisting morbidity, and recent surgery [4]. In addition to these factors, it seems likely that clinical variability in radiation sensitivity can be attributed to profound biological differences between patients [8]. It has long been hoped that identification of these biological differences will lead to more individualized therapeutic strategies and greater understanding of the pathology. Previous attempts to develop predictive assays for RIF have mainly focused on in vitro radiosensitivity and DNA damage endpoints. Although some reports have been promising [9,10], it is still an open question whether assays based
solely on cell killing and DNA damage can be developed into clinically useful predictors [11]. Furthermore, these studies do not provide any new knowledge about the biology behind RIF [11]. More recently, a number of reports have shown that clinically relevant variability in RIF may be associated with different genetic variants [12]. These studies have focused on relatively few single nucleotide polymorphisms (SNPs) in a limited number of candidate genes identified from existing knowledge about the molecular pathology of RIF [13], but have provided important evidence to support the idea that genetic variability is important for the risk of RIF in unselected cancer patients [12]. Microarray-based gene expression studies on the presumed target cell, fibroblasts, can provide new knowledge about genes and pathways affected by various types and doses of irradiation [14–19]. Interestingly, one study comparing ionizing radiation from an external source with DNA-incorporated 125I found a 10-fold higher number of
0167-8140/$ - see front matter c 2007 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.radonc.2007.05.001
262
Fibroblast gene expression and fibrosis
genes induced by external c-irradiation, although the amount of DNA double-strand breaks was higher in cells with DNA-incorporated 125I [20]. Using cDNA microarray data from patient-derived fibroblasts before and after in vitro irradiation, we have recently been able to identify genes, unrelated to DNA damage response, which are differentially expressed between patients with either low or high risk of RIF in patient-derived fibroblasts after irradiation [21]. Differential expression was only observed in the irradiated fibroblasts and only when we looked at two groups of patients, very low risk patients compared to very high risk patients. Interestingly, genes that were generally induced by irradiation were all less upregulated – and genes generally repressed were all less downregulated – in the low risk patients. In the present study, the robustness of radiation-induced changes in gene expression in fibroblasts is addressed by using new batches of cells, new irradiations, different RNA extraction and cDNA synthesis procedures, and measuring gene expression by a different technology (real-time PCR). Furthermore, patients with intermediate to high risk are included, in order to test whether there is a linear correlation between the radiation-induced changes in gene expression and the individual risk of RIF. Finally, differential expression patterns are analysed at the fold induction levels, taking into account the differences in background expression levels between patients.
Cell culture, irradiation, RNA extraction and reverse transcription Fibroblasts were expanded and irradiated as described previously with the exception that only the fractionated scheme of 3 · 3.5 Gy in intervals of 24 h was used [21]. RNA was extracted 2 h after the last fractionated dose using the RNeasy Mini Kit (Qiagen) according to the manufacturers details. RNA was quantified using a spectrophotometer (GeneQuant, Pharmacia Biotech) and 2 lg total RNA was used to generate cDNA with the High Capacity cDNA Archive kit (Applied Biosystem) according to the manufacturers instructions.
Quantitative real-time PCR TaqMan gene expression assays (Applied Biosystems) were used to quantify the following transcripts (gene symbol and Assay ID): ARID5B, Hs01381961_m1; CDC6, Hs00154374_m1; CDON, Hs00610901_m1; CTGF, Hs00170014_m1; CXCL12, Hs00171022_m1; DEGS1, Hs00186447_m1; FAP, Hs00189476 _m1; FBLN2, Hs01002063_m1; LMNB2, Hs00383326_m1; LUM, Hs00158940_m1; MT1X, Hs00745167_sH; MXRA5, Hs00377849_m1; PLAGL1, Hs00243030_m1; PMM1, Hs00160 195_m1; SLC1A3, Hs00188193_m1; SOD2, Hs00167309_m1; SOD3, Hs00162090_m1; and WISP2, Hs00180242_m1. For each reaction, cDNA (corresponding to 5 ng total RNA), 1· assay mix and 1· TaqMan Universal PCR Mastermix (AppliedBiosystems) were mixed in a total of 25 ll. All reactions were performed in duplicate on an ABI Prism 7000 Sequence Detector (AppliedBiosystems).
Materials and methods Patient material
Data analysis
The ‘Aarhus post-mastectomy cohort’ consists of 319 breast cancer patients treated with post-mastectomy radiotherapy between 1978 and 1982 [22,23]. The 26 patients included in the present study are from a smaller cohort of 41 patients from whom cultured fibroblasts are available [24,25]. The patients were treated with an anterior photon field to the axillary and infra/supraclavicular regions, a wax bolus covering the surgical scar in the axilla and with an anterior electron field to the chest wall. In every field, the physical dose absorbed at a dosimetric reference depth of 4.1 mm was calculated and radiation doses were converted into the equivalent biological dose of 2 Gy per fraction using a linear-quadratic model [24]. The evaluation of subcutaneous fibrosis was done by superficial palpation in each of the three treatment fields, and the severity of fibrosis was graded on a four point ordinal scale (grades 0–3) [22] identical to that later used in the LENT-SOMA (Late Effects on Normal Tissues – Subjective, Objective, Management and Analytic) scoring system. Excess risk of fibrosis (an estimate of how the response in a patient compares with the average response in the cohort of 41 patients) is defined as the difference between the observed response in each treatment field of the individual patient and the expected response and was calculated in each field from a dose–response curve based on the 41 patients, taking into account the variation in total dose, dose per fraction and followup time for each patient. The average over the three treatment fields was used to obtain a single measure of clinical radiosensitivity [25,26].
Gene expression levels were normalized to PMM1, phosphomannomutase 1, which based on cDNA microarray data is not affected by irradiation [21]. All assays were tested against PMM1 for equal amplification efficiency based on cDNA standard curves (generated from a mixture of cDNAs from various treated and untreated samples). Twelve assays could be analysed using the comparative CT method (CDON, CTGF, DEGS1, FBLN2, LMNB2, LUM, MT1X, MXRA5, PLAGL1, SLC1A3, SOD2, SOD3) whereas the remaining five were analysed using cDNA standard curves. Fold induction levels were calculated by dividing the relative expression of a given gene to PMM1 in the irradiated sample by the relative expression of the same gene to PMM1 in the unirradiated sample. Two-way clustering was performed using complete linkage hierarchical clustering (rana.lbl.gov/EisenSoftware.htm) of geometric mean centred data, and displayed with Java TreeView (jtreeview.sourceforge.net/). Based on corresponding observations of equivalent biological dose and scoring of normal tissue reaction in each radiation field, dose–response curves were established for subcutaneous fibrosis as described previously [24]. Briefly, the observations were grouped into eight groups according to the equivalent biological dose absorbed at the reference depth. Due to the lower number of observations, the observations were grouped into five and four groups when patients were separated into sensitive and resistant groups, respectively. The incidence of either grade 2 (moderate) or 3 (severe) subcutaneous fibrosis was plotted against the mean radiation dose in each of the dose groups.
J. Alsner et al. / Radiotherapy and Oncology 83 (2007) 261–266
Dose–response curves were fitted to the observations using logit analysis. ED50 values (defined as the radiation dose which would on the average be expected to cause a certain normal tissue effect in 50% of the treatment fields) were established for each curve. Differences in radiosensitivity between genotypes were quantified in terms of enhancement ratios at ED50 level with 95% confidence intervals.
Results In previous studies, cDNA microarray analysis was used to evaluate gene expression in irradiated subcutaneous fibroblasts from breast cancer patients treated with adjuvant radiotherapy [19,21]. By SAM (Significance Analysis of Microarrays), 60 genes were identified that were differentially expressed at >2-fold level between two groups of patients with very low and very high risks of RIF, respectively [21]. Also, by PAM (Prediction Analysis of Microarrays), a minimum set of 18 genes was identified that could differentiate between the two extreme groups [21]. In the present study, we initially selected 17 of these genes for further analysis. Ten genes (ARID5B, CDC6, CDON, CXCL12, DEGS1, LUM, MT1X, MXRA5, PLAGL1 and SLC1A3) were from the PAM gene set, and seven genes (CTGF, FAP, FBLN2, LMNB2, SOD2, SOD3 and WISP2) were selected from the genes identified by SAM. First, new batches of fibroblasts from the same previously studied four patients with very low risk and 10 patients with very high risk of RIF [21] were expanded and irradiated according to the same procedures [19]. Gene expression was measured by real-time quantitative PCR using PMM1 as endogenous control [21] before and after irradiation. In accordance with the cDNA array data [19], 12 of the genes were upregulated by fractionated irradiation (ARID5B, CDON, CXCL12, FAP, FBLN2, LUM, MXRA5, PLAGL1, SLC1A3, SOD2, SOD3 and WISP2) and five genes were downregulated (CDC6, CTGF, DEGS1, LMNB2 and MT1X) (data not shown). Also in accordance with the cDNA array data [21], the 12 genes that were induced by irradiation were less upregulated in the group of fibroblasts from very low risk patients compared to the group of very high risk patients, and likewise, the five repressed genes were less downregulated in fibroblasts from very low risk patients (data not shown). By dividing the relative values of each gene to the endogenous control gene (PMM1) in the irradiated samples with the corresponding relative value in the unirradiated samples, a fold induction level was calculated for all 17 genes. In order to select genes with high differential expression levels, the geometric mean of the fold induction levels in the group of very low risk patients versus the group of high risk patients was compared. For 13 genes (CDC6, CDON, CXCL12, FAP, FBLN2, LMNB2, LUM, MT1X, MXRA5, SLC1A3, SOD2, SOD3 and WISP2) the difference in induction levels was more than 2-fold (data not shown), and these genes were selected for further analysis. The 14 very low and very high risk patients are part of a cohort of 41 patients where fibroblasts as well as data on excess risk of RIF are available (see Materials and methods for details). Next, we extended and irradiated fibroblasts
263
from an additional 12 patients with intermediate to high excess risk scores, and measured gene expression levels for the 13 genes. Fig. 1 illustrates the expression levels before and after irradiation, as well as the fold induction levels, of four representative genes that are either upregulated (MXRA5 (Fig. 1a), and FAP (Fig. 1b)) or downregulated (MT1X (Fig. 1c) and LMNB2 (Fig. 1d)) in fibroblasts from these 26 patients. Expression levels and fold induction levels are plotted against excess risk scores with negative values indicating low risk and positive values indicating high risk of RIF. With the possible exception of MXRA5, none of the 13 genes showed any correlations between expression levels in unirradiated samples and excess risk scores. Differential expression was only observed following irradiation and was most pronounced for the fold induction levels. None of the genes showed any linear correlation between excess risk score and fold induction levels. In general, fibroblasts from patients with very low excess risk scores tended to be least affected by irradiation (having fold induction levels close to 1), whereas the majority of patients with intermediate to high excess risk scores were induced (or repressed) to higher (or lower), but similar, levels. To perform hierarchical clustering analysis of the fold induction levels (Fig. 2), the fold induction value for each gene was divided by the geometric mean value of all fold induction values for that gene. Thus, for the upper gene cluster, which include genes that are generally repressed by irradiation, high values (red) indicate samples with fold induction values closer to 1.0, i.e. less downregulated by irradiation. The lower gene cluster, which include genes that are generally induced by irradiation, low values (green) indicate samples with fold induction values closer to 1.0, i.e. less upregulated by irradiation. Two distinct clusters of patients were identified. The small cluster to the right containing seven patients include the four most resistant patients with the lowest excess risk scores. A dose–response curve was established for the risk of moderate or severe fibrosis in all 26 patients (Fig. 3, left). Patients were then divided into a sensitive and resistant group, respectively, based on the hierarchical clustering analysis in Fig. 2. From the resulting ED50 values (Fig. 3, right), an enhancement ratio of 1.25 (1.12–1.40, 95% confidence interval) was calculated.
Discussion The data presented here first of all demonstrate the robustness of radiation-induced changes in gene expression in fibroblasts. Using new batches of irradiated cells, different RNA extraction and cDNA synthesis procedures, and real-time PCR to measure gene expression, we initially attempted to reproduce expression data generated by cDNA microarrays [21]. In fibroblasts from two groups of patients with either very low or very high risk of RIF, respectively, a perfect correlation was observed between the overall effect of fractionated irradiation (induction or repression) [19]. The 17 genes analysed were selected as they were shown to be differentially expressed between the two groups with
264
Fibroblast gene expression and fibrosis
a
8.00 4.00
16.00 MXRA5
Before After
2.00 1.00 0.50 0.25 0.13 0.06
4.00 2.00 1.00
0.03 0.02 -200 -150 -100 -50
0
0.50 -200 -150 -100 -50
50 100 150 200
Excess risk score
b
4.00
2.00 1.00 0.50
2.00
1.00
0.25 0.13 0
0.50 -200 -150 -100 -50
50 100 150 200
Excess risk score
50 100 150 200
4.00 MT1X
Before After
MT1X
Difference 2.00
2.00 Fold induction
Fold difference relative to PMM1
4.00
0
Excess risk score
8.00
1.00 0.50 0.25
1.00 0.50 0.25
0.13 0.13
0.06 0.03 -200 -150 -100 -50
0
0.06 -200 -150 -100 -50
50 100 150 200
Excess risk score
16.00
0
50 100 150 200
Excess risk score 4.00
32.00 LMNB2
Before After
LMNB2
Difference 2.00
8.00 Fold induction
Fold difference relative to PMM1
FAP
Difference
4.00
0.06 -200 -150 -100 -50
d
50 100 150 200
8.00 FAP
Before After
Fold induction
Fold difference relative to PMM1
8.00
0
Excess risk score
16.00
c
MXRA5
Difference 8.00 Fold induction
Fold difference relative to PMM1
16.00
4.00 2.00 1.00
1.00 0.50 0.25
0.50 0.13
0.25 0.13 -200 -150 -100 -50
0
50 100 150 200
Excess risk score
0.06 -200 -150 -100 -50
0
50 100 150 200
Excess risk score
Fig. 1. Fibroblast gene expression levels before (s) or after 3 · 3.5 Gy/3 days (d) (left), and fold induction levels ( ) (right), for MXRA5 (a), FAP (b), MT1X (c) and LMNB2 (d) plotted against excess risk scores – see text for details.
the induced genes being less induced (and repressed genes being less repressed) in fibroblasts from the resistant group of patients [21]. All 17 genes showed a similar pattern of differential expression, although the degree of differential expression did vary slightly, with some genes showing higher
and some genes showing lower levels of differential expression. Also in accordance with the cDNA data, the expression levels before irradiation varied between patients, but did not show any differential expression between very low and very high risk patients.
J. Alsner et al. / Radiotherapy and Oncology 83 (2007) 261–266
265
Fig. 2. Hierarchical clustering analysis of the fold induction levels relative to the geometric mean values. Patients are identified by the excess risk score as in Fig. 1. 100 Moderate/severe fibrosis (%)
Moderate/severe fibrosis (%)
100 80 60 40 20 0
80 60
Enhancement ratio 1.25 (1.12-1.40)
40 20 0
30
40
50
60
70
80
Radiation dose (Gy) (equivalent dose of 2 Gy per fraction)
30
40
50
60
70
80
Radiation dose (Gy) (equivalent dose of 2 Gy per fraction)
Fig. 3. Dose–response curves for subcutaneous fibrosis in all patients ( ) or divided into groups of sensitive (d) and resistant (s) patients based on gene expression profiles. The enhancement ratio is the ratio between ED50 values (defined as the radiation dose which on average is expected to cause moderate or severe fibrosis in 50% of the treatment fields) with 95% confidence intervals.
Having established the robustness of the assay, 13 genes with the most distinct patterns of differential expression were selected for further analysis. As the previous study only concerned two extreme groups of patients, we next analysed expression patterns in 12 additional patients with intermediate to high risk of RIF. When plotting data before and after irradiation together with the fold induction level (‘after’ value divided by ‘before’ value for each patient), the most pronounced differential expression was observed for the fold induction levels (Fig. 1). It is not known whether the variations in ‘before’ expression are caused by slight differences in experimental setup (e.g. cell density) or by differences in genetic background between patients. However, the effect of fractionated irradiation does not seem to be related to background expression levels, and future studies on any differential effect of irradiation should include background expression measurements. Finally, we could not detect any linear correlation between individual risk of RIF and differential expression of the genes investigated. Rather, a cluster analysis suggests that patients can be divided in two groups with distinct expression patterns. The smaller group contains the four
most resistant patients, as defined by their excess risk score for RIF, together with a few patients with intermediate to high excess risk scores (Fig. 2). Although these latter patients are classified as more or less sensitive by their excess risk score, a dose–response curve (Fig. 3) shows that a cluster analysis based on differential gene expression is capable of dividing patients into two clearly separated groups, a larger, sensitive group and a smaller resistant group. The gene signature presented here does not necessarily represent the most optimal signature for predicting resistance to RIF. Further and larger studies are clearly needed before such a profile can be identified and tested for clinical accuracy. However, with the current experimental setup including isolation and expansion of patient-derived fibroblasts, radiation-induced changes of gene expression in vitro is not an assay that is likely soon to be included in any treatment planning phase. The potential of these studies is rather as an important tool in the identification of the differences in genetic background which leads to the differential induction and repression of genes and pathways between patients with variable risk of RIF, and in the identification of new targets for prevention and intervention of the fibrotic process.
266
Fibroblast gene expression and fibrosis
* Corresponding author. Jan Alsner, Department of Experimental Clinical Oncology, Aarhus University Hospital, Denmark. E-mail address: [email protected] Received 25 March 2007; received in revised form 1 May 2007; accepted 1 May 2007; Available online 23 May 2007
References [1] O’Sullivan B, Levin W. Late radiation-related fibrosis: pathogenesis, manifestations, and current management. Semin Radiat Oncol 2003;13:274–89. [2] Gothard L, Cornes P, Brooker S, et al. Phase II study of vitamin E and pentoxifylline in patients with late side effects of pelvic radiotherapy. Radiother Oncol 2005;75:334–41. [3] Haddad P, Kalaghchi B, mouzegar-Hashemi F. Pentoxifylline and vitamin E combination for superficial radiation-induced fibrosis: a phase II clinical trial. Radiother Oncol 2005;77:324–6. [4] Bentzen SM, Overgaard J. Patient-to-patient variability in the expression of radiation-induced normal tissue injury. Semin Radiat Oncol 1994;4:68–80. [5] Bentzen SM, Christensen JJ, Overgaard J, Overgaard M. Some methodological problems in estimating radiobiological parameters from clinical data. Alpha/beta ratios and electron RBE for cutaneous reactions in patients treated with postmastectomy radiotherapy. Acta Oncol 1988;27:105–16. [6] Kurtz JM. The clinical radiobiology of breast cancer radiotherapy. Radiother Oncol 2005;75:6–8. [7] Yarnold J, Ashton A, Bliss J, et al. Fractionation sensitivity and dose response of late adverse effects in the breast after radiotherapy for early breast cancer: long-term results of a randomised trial. Radiother Oncol 2005;75:9–17. [8] Andreassen CN, Alsner J, Overgaard J. Does variability in normal tissue reactions after radiotherapy have a genetic basis – where and how to look for it? Radiother Oncol 2002;64:131–40. [9] Borgmann K, Roper B, El-Awady R, et al. Indicators of late normal tissue response after radiotherapy for head and neck cancer: fibroblasts, lymphocytes, genetics, DNA repair, and chromosome aberrations. Radiother Oncol 2002;64:141–52. [10] Hoeller U, Borgmann K, Bonacker M, et al. Individual radiosensitivity measured with lymphocytes may be used to predict the risk of fibrosis after radiotherapy for breast cancer. Radiother Oncol 2003;69:137–44. [11] Russell NS, Begg AC. Predictive assays for normal tissue damage. Radiother Oncol 2002;64:125–9. [12] Andreassen CN. Can risk of radiotherapy-induced normal tissue complications be predicted from genetic profiles? Acta Oncol 2005;44:801–15. [13] Bentzen SM. Preventing or reducing late side effects of radiation therapy: radiobiology meets molecular pathology. Nat Rev Cancer 2006;6:702–13.
[14] Ding LH, Shingyoji M, Chen F, et al. Gene expression profiles of normal human fibroblasts after exposure to ionizing radiation: a comparative study of low and high doses. Radiat Res 2005;164:17–26. [15] Tachiiri S, Katagiri T, Tsunoda T, Oya N, Hiraoka M, Nakamura Y. Analysis of gene-expression profiles after gamma irradiation of normal human fibroblasts. Int J Radiat Oncol Biol Phys 2006;64:272–9. [16] Ishikawa K, Koyama-Saegusa K, Otsuka Y, et al. Gene expression profile changes correlating with radioresistance in human cell lines. Int J Radiat Oncol Biol Phys 2006;65:234–45. [17] Kis E, Szatmari T, Keszei M, et al. Microarray analysis of radiation response genes in primary human fibroblasts. Int J Radiat Oncol Biol Phys 2006;66:1506–14. [18] Heinloth AN, Shackelford RE, Innes CL, et al. Identification of distinct and common gene expression changes after oxidative stress and gamma and ultraviolet radiation. Mol Carcinog 2003;37:65–82. [19] Rødningen OK, Overgaard J, Alsner J, Hastie T, BørresenDale AL. Microarray analysis of the transcriptional response to single or multiple doses of ionizing radiation in human subcutaneous fibroblasts. Radiother Oncol 2005;77:231–40. [20] Sokolov MV, Smirnova NA, Camerini-Otero RD, Neumann RD, Panyutin IG. Microarray analysis of differentially expressed genes after exposure of normal human fibroblasts to ionizing radiation from an external source and from DNA-incorporated iodine-125 radionuclide. Gene 2006;382:47–56. [21] Rødningen OK, Alsner J, Hastie T, Overgaard J, Børresen-Dale AL. Application of microarray analyses to identify genes involved in radiation-induced fibrosis. Breast Cancer Res 2005;7(Suppl 2):542–3. [22] Overgaard M, Bentzen SM, Christensen JJ, Madsen EH. The value of the NSD formula in equation of acute and late radiation complications in normal tissue following 2 and 5 fractions per week in breast cancer patients treated with postmastectomy irradiation. Radiother Oncol 1987;9:1–11. [23] Bentzen SM, Overgaard M, Thames HD, Christensen JJ, Overgaard J. Early and late normal-tissue injury after postmastectomy radiotherapy alone or combined with chemotherapy. Int J Radiat Biol 1989;56:711–5. [24] Andreassen CN, Alsner J, Overgaard M, Overgaard J. Prediction of normal tissue radiosensitivity from polymorphisms in candidate genes. Radiother Oncol 2003;69:127–35. [25] Johansen J, Bentzen SM, Overgaard J, Overgaard M. Relationship between the in vitro radiosensitivity of skin fibroblasts and the expression of subcutaneous fibrosis, telangiectasia, and skin erythema after radiotherapy. Radiother Oncol 1996;40:101–9. [26] Bentzen SM, Overgaard M, Overgaard J. Clinical correlations between late normal tissue endpoints after radiotherapy: implications for predictive assays of radiosensitivity. Eur J Cancer 1993;29A:1373–6.