The Relationship Between Homologous Recombination Repair and the Sensitivity of Human Epidermis to the Size of Daily Doses Over a 5-Week Course of Breast Radiotherapy

This article is featured in Highlights of This Issue, p. 5155

Clinical radiotherapy is delivered as a sequence of fractional, usually daily, doses. Tissues vary in their sensitivity to several treatment-related variables, including total dose, fraction size, interfraction interval, and overall treatment time (1). On average, cancers are less sensitive to fraction size than the normal tissues responsible for most dose-limiting complications months or years later (2). High total doses (>60 Gy) delivered in small (2.0 Gy) fractions spare normal tissues relative to most malignancies, and achieve the highest level of tumor control for a given level of chronic adverse effects. Recent level 1 evidence indicates that breast and prostate cancer are more sensitive to fraction size than commonly assumed (3–5). An understanding of molecular mechanisms is, therefore, needed to ensure effective selection of fractionation schedule in individual patients.

Fraction size sensitivity is a cellular property reflecting, the ability to repair otherwise lethal DNA double-strand breaks (DSB) before the next fraction of radiotherapy (6). DSB are rapidly repaired by nonhomologous end-joining (NHEJ) in all phases of the cell cycle, in addition to which, homologous recombination requiring an intact sister chromatid, repairs a proportion of DSB in S and G2 phases of the cell cycle (7–9). Fractionation sensitivity is associated with the proliferative status of a tissue, being higher in tissues with low-proliferative indices, and vice versa (10). Rodent cell lines deficient in NHEJ are very radiation sensitive, relying disproportionately on homologous recombination to repair DSB and showing no dose-rate sparing, an indicator of insensitivity to fraction size (11). We postulated that the high fidelity of DSB repair in replicated chromatin (12, 13), which is independent of DNA damage levels, may explain the low-fractionation sensitivity of tissues with high-proliferative indices, including self-renewal normal tissues and many cancers.

The epidermis in patients undergoing radiotherapy for breast cancer offers a readily accessible system for investigating DSB repair in different proliferative states. Early in the course of treatment, clinical data suggest that the epidermis is more sensitive to fraction size than at the end of a 5-week schedule of radiotherapy as showed by Turesson and colleagues in breast cancer patients irradiated with various dose schedules (10, 14). Before radiotherapy, basal epidermal cells are predominantly in nonproliferative phase, but proliferative markers and cell cycle checkpoints change over the course of 5 weeks radiotherapy (15, 16). DSB rejoining can be monitored in tissue sections using antibodies against 53BP1 as a surrogate for DSB and antibodies against RAD51 as a biomarker of homologous recombination (17).

A total of 30 patients prescribed radiotherapy to a dose of 50 Gy in 25 fractions (2.0 Gy per fraction) over 5 weeks to the breast after tumor excision of early breast cancer were recruited with written informed consent. Radiotherapy was delivered to the whole breast using tangential 6 to 10 MV X-ray beams. We adhered to the International Commision on Radiation Units and Measurements (ICRU) recommendations to keep dose inhomogeneity between 95% and 105% in the target volume and three-dimensional dose compensation was used to achieve homogeneity. The estimated dose to the skin was 1.6 Gy per fraction. Ethical and scientific approval for the study was granted by the Royal Marsden Hospital Research Ethics Committee (REC No: 06/Q0801/73) and Committee for Clinical Research (CCR 2900).

Single 4 mm punch biopsies of breast skin were collected at the following time points and locations:

• 2 hours (h) after the 1st fraction from irradiated and contralateral breast.

• 2 hours after the 5th fraction from irradiated breast.

• 1 hour before and 2 hours after the final (25th) fraction from irradiated breast.

Local anesthetic (0.5–1 mL lidocaine) was infiltrated subcutaneously before biopsies collected at least 2 cm inside the lateral border of the breast volume and at least 1 cm apart. Steri-strips plus dry nonadhesive gauze were applied to approximate the wound edges. Biopsies were fixed in 10% neutral buffered formalin, embedded in paraffin and cut into 4 μm sections. Four punch biopsies, 2 from the irradiated breast and 2 from the unirradiated breast, were collected immediately after the 25th fraction of radiotherapy in 3 patients. One paired sample from each of the 3 patients was snap frozen in liquid nitrogen and the other fixed in formalin. No problems with healing or infection were encountered.

To avoid loss of tissue section adherence following antigen retrieval, all sections were picked up onto either superfrost plus or superfrost gold slides and placed on a hot plate at 60°C for 10 minutes before starting. Deparaffinized and rehydrated samples were then microwaved at 850 W in 250 mL of 10 mmol/L citric acid (pH adjusted to 6.0) for 12 minutes (3 × 4 minutes). They were then left to stand at room temperature for 20 minutes, before being washed in running tap water, rinsed in TBS, and the section circled with a resin pen. We conducted double staining in layers as described before (18; for all antibody details please refer to Table 1). Briefly, after treating with DAKO peroxidase and protein block for 5 minutes each, the sections were incubated with β1 integrin (CD29) for 1 hour at room temperature. After washing with TBS, mouse/rabbit Dako EnVision horseradish peroxidase (HRP) was applied for 30 minutes. Following further washing in TBS, diaminobenzadine (DAB) solution was applied for 2 minutes. For the second layer of staining, 250 mL of 10 mmol/L citric acid (pH adjusted to 6.0) was brought to the boil by microwaving at 850 W for 4 minutes and the slides immersed in it for 1 to 2 minutes (to remove any unbound antibody) before cooling down to room temperature. Dako protein block was again applied for 5 minutes, and the sections were incubated for 1 hour at room temperature with Ki67 (MIB-1) or 53BP1 antibody. Similar to the first layer, mouse/rabbit Dako EnVision HRP was applied for 30 minutes followed by Vector SG (Vector laboratories) for 5 to 10 minutes. Counterstaining was conducted using methyl green (Surgipath) for 20 seconds followed by dehydration, clearing in xylene and mounting in DPX (a standard mounting media in histology containing distyrene, a plasticizer, and xylene).

Cyclin A and p21 single staining was done on serial 4 μm sections of epidermis as described above for the first layer. To compare β1 integrin staining with melanoma chondroitin sulphate proteoglycan (MCSP), which works best on fresh frozen skin samples, we obtained paired biopsies from 3 patients, one of which was fixed in formalin and the other snap frozen. For MCSP staining, the snap frozen skin biopsies were embedded in cryogel, sectioned and fixed in cold acetone for 10 minutes. No antigen retrieval was required. After peroxidase and protein block, the sections were incubated with the MCSP antibody for 60 minutes at room temperature followed by EnVision HRP and DAB as described above. Haematoxylin was used as the counterstain.

For double fluorescent staining, the sections were deparaffinized, rehydrated, and heat-mediated antigen retrieval carried out as above. After 5 minutes in Dako protein block, the sections were incubated with RAD51 (51RAD01) and Ki67 (NCL-Ki67p) antibody for 60 minutes at room temperature. After washing with 3% fetal calf serum (FCS) in PBS, Alexa Fluor 488-conjugated goat anti-mouse immunoglobulin G (IgG; Invitrogen, 1:200) and TRITC-conjugated donkey anti-rabbit IgG (1:200) was applied for 60 minutes in the dark at room temperature. This was followed by washing with 3% FCS and staining with 4′, 6-diamidino-2-phenylindole (DAPI; 1:200) for 5 minutes. The slides were air dried and mounted in Vectashield. Double staining with 53BP1 and Ki67 (MIB-1) as well as Ki67 (NCL-Ki67p) and cyclin A was done in a similar fashion. Secondary antibody combinations (conjugated with Alexa Fluor 488 or 555, all at 1:200 dilutions) were applied as appropriate.

The entire interfollicular basal compartment was analyzed in each section. Analysis of the immunohistochemical data was undertaken by visual inspection and counting of the slides at high power (X60) and was facilitated for some of the analyses by capturing digital images using an Axioscope trans-illumination microscope (Zeiss) connected to a 3 charged-coupled devices (CCD) color camera (JVC) connected to a PC with a Matrox Meteor frame grabber in a peripheral component interconnect (PCI) bus. The basement membrane length was measured using a calibrated line tool developed in Visilog 5.02 software (Noesis Vision Inc.). β1 integrin and p21 staining was scored as negative (–), weak (+), moderate (++), or strong (+++) depending on intensity of staining. Only cells that were attached to the basement membrane and had moderate to strong uptake of β1 integrin along the entire cell membrane were considered to be positive (putative stem cells). Any nuclear uptake of Ki67 (positive in all phases of cell cycle except G0) or cyclin A (S-G2 phase marker) was scored as positive.

All analysis was done manually at 100× in oil using a Nikon Eclipse TE200 epifluorescence microscope. Digital images were obtained with a cooled analog camera via a video frame grabber and PC. To study the DNA DSB repair pathways in the proliferating versus nonproliferating cell population of the basal epidermis, we looked at RAD51 and 53BP1 foci numbers in all Ki67 positive versus negative cells along the basement membrane. To assess what proportion of proliferating cells were in S-G2 phase of the cell cycle, all Ki67⁺ cells in the section were analyzed for cyclin A uptake by costaining.

Paired Student t test (2-tailed) was applied to compare the results from the different time points. The level of significance was taken as P < 0.05. One star refers to a P value of <0.05, two stars P < 0.01, and three stars P < 0.001.

The mean age of patients recruited was 58 years. Three patients experienced grade 3 erythema and 3 had moist desquamation in the inframammary fold toward the end of radiotherapy. None of the patients recorded severe late toxicities after 5 years of follow-up. To our knowledge, none of these patients had BRCA1/2-associated familial cancers.

A majority of basal cells showed changes in Ki67 tissue immunohistochemistry (IHC) during the course of radiotherapy (Fig. 1A). The mean basal Ki67 density (cells per mm of basement membrane) increased from 5.72 to 15.46 per mm of basement membrane by the end of the week 5 (P = 0.002; Fig. 1B). Immunostaining for cyclin A protein identified the proportion of cells in the late S and G2 phases of the cell cycle. Basal cyclin A index (number of positive cells/total number of cells in basal layer) rose from 3.7% at baseline to 12.8% at 5 weeks. To accurately analyze cell cycle stage, we costained with Ki67 and cyclin A (Fig. 1A), revealing a significantly higher proportion of cycling cells in S or G2 phases of the cell cycle by the end of week 1 of radiotherapy (48%) as compared with baseline (29%), and the increase was maintained at the end of 5 weeks [P < 0.0003, 95% confidence interval (CI) 47.7–50.2; Fig. 1C]. These findings are consistent with other reports of human epidermis in response to fractionated radiotherapy, detected using single staining with Ki67 and cyclin A (15).

It is well established that ionizing radiation (IR) triggers cell cycle arrest, and it was initially surprising to observe a higher proportion of cycling cells after 5 weeks of radiotherapy. One explanation is that cells accumulate in cell cycle phases with enhanced capacity to survive IR-induced damage. We immunostained for the cyclin-dependent kinase inhibitor p21, which activates several DNA damage-induced cell cycle checkpoints (19). In our study, there was virtually no expression of p21 at baseline, and only 2.8% of cells showed mild (+) staining 2 hours after the first fraction of radiotherapy (95% CI 2–3.5). The p21 index subsequently rose to 87.4% by the end of 5 weeks of radiotherapy (P < 0.0001, 95% CI 84.1–90.6; Fig. 2A and B). When serial sections stained for cyclin A and p21 were overlaid using MATLAB 7.9 R2009b software, the cyclin A positive nuclei overlapped with strong p21 uptake (Supplementary Fig. S1).

Ki67 and cyclin A staining was seen only in the basal or first suprabasal layer, but nuclear staining of p21 was seen dispersed in all layers of the epidermis with strongest uptake in the basal compartment. The upregulated p21 in Ki67 negative (i.e., G0) cells, coupled with a decrease in the proportion of G1 cells is likely to be related to a radiation-induced G0 block as suggested by Turesson and colleagues (15). Our data suggests that although a significantly high proportion of basal cells enter the cell cycle in response to radiation-induced cell depletion, effective cell cycle checkpoint activation prevents them from entering mitosis while radiotherapy is ongoing. Staining with the mitosis marker phospho-Histone H3 does not show an increase until after the end of a 5-week course of radiotherapy (16).

Next, we stained for RAD51 foci formation as a measure of active homologous recombination (17). As expected, RAD51 foci were seen only in Ki67 positive cells (Fig. 3A), consistent with homologous recombination activation in S/G2 phase cells (7–9). Ki67 positive cells arrested in S/G2 phase are able to trigger Chk1 DNA damage response (DDR) and homologous recombination repair (20). Dual staining with Ki67 helped to identify RAD51 foci more easily as this protein is expressed only in cycling cells (21). RAD51 foci have been previously visualized in human tissue after ex vivo radiation (22) and in vivo chemotherapy (23). However, to our knowledge this is the first demonstration of RAD51 foci formation in human tissue following in vivo radiotherapy. The inherent background noise in tissue sections needed to be optimized to accomplish this (Fig. 3A).

The number of Ki67 positive nuclei in the basal epidermis containing RAD51 foci increased 3.9-fold between day 1 and 33 (range 2.6–7.2, P = 0.0003; Fig. 3B). Taking into account the increased number of Ki67 positive cells during the course of fractionated radiotherapy, the total RAD51 positive cell density (per millimeter of basement membrane) increased 9.7-fold (Fig. 3C). The mean RAD51 foci count per foci-containing cell 2 hours after the first fraction was 2.5 compared with 3.5 after the last fraction of radiotherapy (P = 0.01).

It is well established that cells present in the late S/G2 phases are more resistant to IR as compared with other phases, pointing to homologous recombination as a possible mechanism for minimizing residual DSBs (24). To analyze DSB repair in cycling versus noncycling cells during 5 weeks of fractionated radiotherapy, we costained with Ki67 and 53BP1 (Fig. 4A). At baseline (unirradiated controls) there was no significant difference in the number of 53BP1 foci seen in Ki67 positive (mean 0.6) versus Ki67 negative cells (mean 0.7, P = 0.1; Fig. 4B). Interestingly, the mean number of residual foci remaining just before the last fraction of radiotherapy was 1.6 and 2.0 in Ki67 positive and negative cells, respectively (P = 0.01). Overall, there was a 2.6-fold (P = 0.001) increase in the number of residual 53BP1 foci remaining before the last fraction. However, it is not possible to say if residual foci just before the last fraction represented an accumulation of unrepaired DSB over weeks, or residual foci from the previous day's dose.

When examining 53BP1 foci 2 hours after the first fraction of radiotherapy, we found that the average foci numbers were significantly higher in Ki67 negative (5.82) versus Ki67 positive (3.48) nuclei (P < 0.0001; Fig. 4C). A similar finding was seen 2 hours after the last fraction with Ki67 negative cells having a mean foci number of 7.24 as compared with 4.59 in Ki67 positive cells (P = 0.0002; Fig. 4C). In the Ki67 positive population around 20% to 30% of nuclei failed to show distinct 53BP1 foci, but rather a diffuse staining pattern (similar to unirradiated nuclei) suggesting completion of DSB repair. Cells with diffuse staining pattern were scored as negative. This diffuse staining pattern was not observed in any of the Ki67 negative cells counted along the basal layer 2 hours after treatment.

The human epidermis consists of quiescent stem cells and cycling progenitors/transit-amplifying cells that amplify the number of differentiated daughters resulting from each stem cell division (25). Several research groups have made extensive efforts to define a set of stem cell-specific markers in the epidermis. Much of this research has used a combination of cell adhesion molecules such as α6 or β1 integrin and the absence of cell proliferation markers such as Ki67 (25–27). In our study a combination of β1 integrin and Ki67 were used to identify putative epidermal stem cells (β1integrin ++/+++ and Ki67 negative). At baseline in unirradiated skin sections, only 20% of proliferating cells were β1 integrin positive, whereas 84% were β1 integrin positive by the end of 5 weeks of radiotherapy (P < 0.00001; Fig. 5). By week 5, the distinct dermal papillae had disappeared and the clustered pattern of β1 integrin staining was lost. Not only did the fraction of basal cells expressing β1 integrin increase dramatically from 5% to 10% at baseline to around 85% at week 5, the intensity of staining increased as well (Fig. 5A). Using a more specific stem cell marker MCSP (28), we found a similar change in staining pattern with radiotherapy. In the snap frozen sections, at baseline, MCSP uptake was highly clustered in a few basal cells (3%–5%), but became diffuse after 5 weeks of radiotherapy, with most of the basal layer cells staining positive (data not shown).

Clinical data suggest that sensitivity of epidermis to fraction size varies over a 5-week course of radiotherapy, being sensitive to fraction size at the beginning of radiotherapy, and losing sensitivity during weeks 4 and 5 (10, 14). Our data suggests that activation of homologous recombination to repair DSB toward the end of radiotherapy correlates with this loss of fractionation sensitivity seen clinically. NHEJ is an error prone method of repair although homologous recombination has a high fidelity for repair that is independent of DNA damage levels (fraction size; 7). Another mechanism contributing to loss of fraction size sensitivity in S/G2 could be via cohesin, which facilitates more efficient DSB repair in replicated chromatin (29) and may enhance correct DSB repair at high doses in S/G2 in an homologous recombination-independent manner (13). The tightly bound chromatid in S/G2 might serve as a scaffold to hold break ends in place and thus avoid loss or rearrangement of chromosome material. Without such a scaffold (i.e., in G0/G1), break ends would be able to move more freely, repair would be slower and more frequently result in “wrong” ends being joined, especially at higher doses (or dose per fraction) when breaks are in close proximity. Whatever mechanisms are involved, cells in S-G2 appear to have a high fidelity for repair that is independent of DNA damage levels, explaining insensitivity to fraction size and mediating resistance to radiation (7, 30). Thus, a higher proportion of cells in S/G2 phase of the cell cycle (allowing greater use of high-fidelity DSB repair) is more likely to influence fractionation sensitivity than proliferation per se (as determined by Ki67 staining).

The 53BP1 protein accumulates in discrete nuclear foci at DSB (31), which can be used as a marker for DSB. We have been cautious in not interpreting the RAD51 and 53BP1 data in terms of relative use of homologous recombination and NHEJ for several reasons. It was technically challenging to triple stain with Ki67, RAD51, and 53BP1 and the 2-hour time point may underestimate NHEJ repair. We have therefore used 53BP1 to study the total remaining DSB and found that the number of residual DSB is lower in Ki67 positive cells as compared with Ki67 negative cells 2 hours after radiotherapy (Fig. 4C). This is surprising, as it previously has been shown that cells in G2 acquire almost twice as many DSB as those in G1 after a 2 Gy IR treatment, which is explained by the increased DNA content in G2 cells (7, 9). Thus, our finding that there are overall fewer 53BP1 foci in Ki67 positive cells suggests an increased DSB repair capacity of late S/G2 cells, probably obtained by activated homologous recombination. This is consistent with the findings of Rothkamm and colleagues in rodent cell lines (7). Residual 53BP1 foci observed before the last fraction, following 5 weeks of radiotherapy (Fig. 4B) may not just reflect unrepaired damage directly induced by radiation. It is possible that stress/senescence associated 53BP1 foci may contribute (32). Recent evidence suggests a preferential association of such foci with telomeres (33).

Supiot and colleagues have shown upregulation of p21 with a decreased Ki67 after 5 days of prostate radiotherapy suggesting that terminal growth arrest rather than apoptosis maybe the dominant mode of radiation-induced cell death in prostate epithelium (34). In our study, after the initial drop in Ki67 staining after 5 days there was a significant increase in the Ki67 staining in the basal epithelium after 5 weeks of continuous radiotherapy. This was accompanied by increased p21 staining which indicates not just G1 arrest but also G2 arrest as shown by overlaying serial sections stained for cyclin A and p21 (Supplementary Fig. S1). As shown by Turesson and coleagues, the mitosis marker phospho-Histone H3 starts to increase significantly in the basal epithelial cells after the end of radiotherapy (16). Thus, it appears that these basal epithelial cells are likely to be composed of stem/progenitor cells that have not undergone permanent cell cycle arrest/senescence. The increased p21 reflects checkpoint activation preferably in S-G2, which allows the cells to repair their DSB most efficiently. After radiotherapy these basal cells proliferate to replenish the epithelium. Our study suggests that a high proportion of cells in (or arrested in) S-G2 may influence the fractionation sensitivity of the tissue as a whole.

It is suggested that normal and cancer stem cells display a radioresistant phenotype due to activation of a DDR that enhances repair of radiation-induced DNA damage (35). The hypothesis is that stem cells make up an increasing proportion of surviving cells after each fraction of radiotherapy. In this study we found that the number of cells positive for the putative epidermal stem cell marker β1 integrin (25, 26, 27), increased during radiotherapy, in line with the notion that stem cells are radioresistant (35, 36; Fig. 5). The β1 integrin positive cells were to a high degree also Ki67 positive, suggesting that many of the stem cells have entered the cell cycle. Altogether, our data may suggest that the cells persisting after 5 weeks of radiotherapy are epidermal stem cells that have arrested in the late S-G2 phase of the cell cycle and activated homologous recombination. However, we remain very uncertain of such a conclusion as our study suggests that integrins and MCSP may not be the best markers to study stem cells in the setting of radiotherapy due to radiation-induced upregulation. Cell extracellular matrix contact mediated via integrins is thought to have an impact on cellular mechanisms resulting in increased cell survival upon exposure to IR (37, 38). Several human tumor cell lines and normal human fibroblastic cell strains have shown to increase β1 integrin expression in a dose-dependent manner following IR (39). We could not find other reports describing this in skin in vivo. It is difficult to comment if β1 integrin positivity after 5 weeks of radiotherapy is representative of true stemness or inherent radioresistance. However, the finding that Ki67 positive cells (the majority of which are β1 integrin positive at 5 weeks) show increased DSB repair capacity, suggest that these are surviving radiotherapy and are likely to play a role in repopulating the epidermis after radiotherapy.

Are the responses to radiotherapy in skin tissue reported here also relevant to cancer? Here, we find that epidermal cells accumulate in late S/G2 phases of the cell cycle, allowing them to trigger homologous recombination. It is widely accepted that normal fibroblast cultures with intact p53 arrest at the G1/S checkpoint, which is thought to be less leaky than the G2 checkpoint. However, there is evidence that human epidermal cells (keratinocytes) may have a differential response compared with fibroblasts, in that they predominantly arrest at the G2 checkpoint following radiation (40). The attenuated G1 arrest in keratinocytes correlated with reduced p53 accumulation compared with G1 arrested fibroblasts (40). Cancer cells often lose cell cycle checkpoints such as the p53 pathway and accumulate in the G2 phase of the cell cycle after DNA damage (41). Thus, it is highly likely that the increase in homologous recombination reported here in skin epithelial cells also extends to many cancer cells, especially those with a high-proliferative index.

In conclusion, although it is known that most of the IR-induced DSB are repaired by NHEJ in mammalian cells, our results suggest that homologous recombination plays an important role in DSB repair in the human epidermis by the end of a 5-week course of fractionated radiotherapy as cells accumulate in S/G2 phase of the cell cycle (Fig. 3D). Adoption of homologous recombination, because of its high fidelity, offers a mechanism explaining loss of fractionation sensitivity in rapidly cycling normal and malignant tissues, and suggests a potential approach to individualization of radiotherapy dose prescription.

No potential conflicts of interest were disclosed.

Conception and design: N. Somaiah, J.R. Yarnold, K. Rothkamm, T. Helleday

Development of methodology: N. Somaiah, J.R. Yarnold, F. Daley, A. Pearson, K. Rothkamm

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Somaiah, A. Pearson, L. Gothard, K. Rothkamm

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Somaiah, K. Rothkamm, T. Helleday

Writing, review, and/or revision of the manuscript: N. Somaiah, J.R. Yarnold, K. Rothkamm, T. Helleday

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N. Somaiah, F. Daley, A. Pearson, L. Gothard, T. Helleday

Study supervision: K. Rothkamm, T. Helleday

This study has been kindly funded by the Breast Cancer Campaign charity, grant reference 2006NovPR11, the Swedish Cancer Society and the Swedish Research Council. We also acknowledge NHS funding to the NIHR Biomedical Research Centre and the NIHR Centre for Research in Health Protection.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.