Inhibition of Homologous Recombination and Promotion of Mutagenic Repair of DNA Double-Strand Breaks Underpins Arabinoside–Nucleoside Analogue Radiosensitization

Concurrent chemoradiotherapy (cCRT) is an essential treatment option for the management of solid tumors. Nucleoside analogues are a class of chemotherapeutic drugs with antitumor activity. Some nucleoside analogues also show strong radiosensitizing potential (1) and are therefore relevant to cCRT. Our early work showed that a NA, 9-β-D-arabinofuranosyladenine (ara-A), has an impressive radiosensitizing potential (2). Indeed, araA strongly inhibits repair of a slow form of potentially lethal damage in plateau phase rodent cells (3, 4), and sensitizes cycling human cells to ionizing radiation (5). The molecular mechanisms underpinning this radiosensitization remain uncharacterized, despite their potential utility in the development of improved compounds with similar mechanism of action, as well as in the development of optimized administration protocols for existing ones.

Several effects of nucleoside analogues on the cellular metabolism are implicated in their mechanism of radiosensitization. These effects include inhibition of DNA synthesis and repair, induction of apoptosis, redistribution of cells in the cell cycle, interference with nucleotide metabolism, and alterations in the balance of intracellular nucleotide pools (1, 6, 7).

A major determinant of cellular radiosensitivity to killing is repair of DNA double-strand breaks (DSB; ref. 8). In mammalian cells, ionizing radiation-induced DSBs are thought to be processed predominantly by the classical, DNA-PK–dependent pathway of nonhomologous end-joining (D-NHEJ), in which the final ligation is carried out by DNA LIGASE IV (9). D-NHEJ is active throughout the cell cycle and operates with fast kinetics (10).

DSBs can also be repaired by alternative pathways of NHEJ functioning as a back-up (B-NHEJ) to other DSB repair pathways. Such backup functions may prevail either when other DSB repair pathways are genetically and chemically inactivated, or when they fail locally in a repair-proficient cell. It is commonly assumed that B-NHEJ functions mainly as backup to D-NHEJ. However, recent work from our laboratory provides evidence that B-NHEJ also operates as a backup to homologous recombination (HRR; ref. 11). Components of the MRN (MRE11-RAD50-NBS1) complex as well as CtIP (CtBP-interacting protein), have been implicated in B-NHEJ, and are likely to promote B-NHEJ through their involvement in DNA end resection (12, 13).

In dividing cells, B-NHEJ shows robust activity throughout the cell cycle and maximum efficiency in G₂ (14). B-NHEJ causes more extensive deletions at DSB junctions than D-NHEJ and, importantly, markedly promotes chromosome translocations (15–17). Thus, B-NHEJ is now widely considered as a main source of genomic instability in cells of higher eukaryotes.

Finally, DSBs can be processed by HRR. HRR requires a homologous template–DNA strand to faithfully restore DNA sequence in the vicinity of a DSB. In higher eukaryotes, the preferred donor sequence is the sister chromatid, which restricts HRR to the S- and G₂ phases of the cell cycle (18). Prerequisite for HRR is the resection of DSB ends to create RPA-coated 3′-ssDNA overhangs. The RAD51 recombinase, a key component of HRR apparatus, is loaded on this ssDNA aided by proteins of the RAD52 epistasis group and the RAD51 paralogs (18). RAD51 nucleoprotein filament formation is followed by search for homology and strand exchange that is resolved to give gene conversion, or more rarely crossing over events at the DSB site. HRR has been directly linked to the S-phase–dependent radioresistance to killing and pharmacologic inhibition of HRR has been shown to strongly sensitize cells to ionizing radiation (19, 20).

Recombination can also occur between homologous sequences flanking a DSB on the same DNA molecule. This type of repair is termed single strand annealing (SSA) and is a mutagenic mode of homology-directed repair as the intervening DNA sequence is lost (21). SSA is strongly suppressed when a functional HRR apparatus is active in the cell (22, 23).

The quantitative contribution of HRR to the overall processing of ionizing radiation induced DSBs is still debated, but is commonly regarded to be much smaller than that of D-NHEJ (24, 25). Defects in HRR and D-NHEJ lead to radiosensitization (26, 27). The role of B-NHEJ in the radiosensitivity of repair-proficient, or repair-deficient cells remains to be elucidated.

Here we investigate the effects of araA and other nucleoside analogues on DSB repair in human cancer cells exposed to ionizing radiation. Our results demonstrate a previously underappreciated potential of araA and other nucleoside analogues, to inhibit HRR and to promote mutagenic DSB repair.

A549 human lung adenocarcinoma cells and U2OS human osteosarcoma cells carrying different reporter substrates (282C/DR-GFP, 280A/EJ5-GFP, and 283C/SA-GFP, a gift of Dr. Jeremy Stark, Beckman Research Institute of City of Hope, Duarte, CA and U2OS EJDR cells, a gift of Dr. Simon Powell, Memorial Sloan-Kettering Cancer Center, New York, NY) were maintained in McCoy 5A medium. The reporter constructs integrated in U2OS cells have been described (28, 29). DRaa40 cells were grown in DMEM. All cell lines were cultured with 10% FBS at 37°C and 5% CO₂. Human cell lines were subcultured twice, CHO cells thrice a week. A549 and U2OS-282C cells were authenticated using Multiplex Cell Authentication by Multiplexion as described (30). All other U2OS reporter cell lines could be unequivocally identified on the basis of their integrated reporter constructs.

For experiments in the plateau phase of growth, A549 cells were plated as for regular subculture, but grown for 7 days without medium change; at that time, the cell number reached a plateau and cells accumulated (∼90%) with a G₁ DNA content (Supplementary Fig. S1). All experiments were carried out under normoxic conditions.

Cell transfections were carried out with a Nucleofector device (Lonza). Drugs were added at the required concentrations 1.5 hours after transfection and incubation was carried out for 4 hours. Flow cytometry was carried out with a Beckman Coulter XL-MCL or a Beckman-Coulter Gallios flow cytometer. Cells were analyzed 24 hours after transfection, unless stated otherwise.

Nucleoside analogues were prepared and frozen as 100 mmol/L stocks in the following solvents: araC, H₂O; araA, acidified water (0.2 mol/L HCl); F-araA, DMSO. Hydroxyurea (HU) stocks were prepared at the same concentration in H₂O. Aphidicolin (Aph) was prepared at 200 μg/mL in DMSO.

Cell irradiations were performed using an X-ray machine (GE Healthcare) operated at 320 kV, 10 mA with a 1.6 mm aluminum filter. The dose rate at 50 cm was approximately 2.8 Gy/minute.

Transfection of siRNA was performed using the Nucleofector device (Lonza). For knockdown of RAD51, we used the FlexiTube-siRNA-Hs_RAD51_7 (Qiagen; Cat. no.: SI02663682) and as negative control siRNA against GFP (GFP-22; Qiagen; cat. no.: 1022064). Knockdown was confirmed by Western blotting with primary antibodies against RAD51 [rPAb RAD51(Ab-1); Calbiochem] and GAPDH (MAB374; Millipore) and goat secondary antibodies labeled with infrared dyes (IRDye-680LT and IRDye-800CW; Li-Cor Biosciences). Imaging was performed with an Odyssey scanner (Li-Cor Biosciences).

Pulsed-field gel electrophoresis (PFGE) of exponentially growing, plateau phase, and sorted cell populations was performed as described (14). Cells were subject to FACS using a Beckman-Coulter Epics Altra sorter. Gels were visualized using a phosphoimager (Typhoon, GE Healthcare) and quantitated using the ImageQuant 5.2 software (GE Healthcare). Results are expressed as dose-equivalent in Gy (DEQ), which is calculated using a dose–response curve generated with the same cells (14).

A549 cells were plated after the appropriate dilution, before irradiation and drug treatment. Drugs were added 20 to 30 minutes before irradiation and left for postirradiation incubation for 4 hours. Cells were washed twice and supplied with fresh growth medium. Colonies were stained with crystal violet and counted 9 to 10 days later.

Cells grown on coverslips were fixed with 2% paraformaldehyde and permeabilized with 0.5% Triton-X100. For detection of 5-ethynyl-2′-deoxyuridine (EdU), cells were processed with the Click-iT-EdU-Alexa-Fluor-647 imaging kit according to manufacturer's instructions. The following primary antibodies were used: mouse monoclonal IgG2b-RAD51-14B4 (GeneTex) and rabbit polyclonal IgG cyclin-B1-H-433 (Santa Cruz Biotechnology). RAD51 was detected using a goat polyclonal Alexa488-conjugated anti-mouse IgG secondary antibody and cyclin B1 using a goat polyclonal Alexa568-conjugated anti-rabbit IgG antibody (Invitrogen). DNA was stained with DAPI and coverslips were mounted on slides with ProLong-Gold anti-fade (Invitrogen) and analyzed using a LEICA-TCS-SP5 system.

We initiated experiments to investigate the effect of selected nucleoside analogues (Fig. 1A) at different endpoints using A549 cells. Figure 1B shows survival of cells exposed postirradiation for 4 hours to the indicated concentrations of araA. AraA strongly radiosensitizes A549 cells producing shoulderless survival curves above 500 μmol/L. From these data, araA sensitizer enhancement ratios were calculated at a fixed dose of 2 Gy, or at a fixed survival of 37%. The results summarized in Supplementary Table S1 document strong radiosensitization.

As araA has relatively limited clinical use, we determined the radiosensitizing potential of F-araA and araC, two nucleoside analogues extensively used in the clinic, using similar experimental protocols. In these experiments, nucleoside analogue concentrations exerting similar toxicity in nonirradiated cells, as determined by plating efficiency analysis, were chosen as follows: Ctrl. 65% ± 4%; 500 μmol/L araA 40% ± 9%; 500 μmol/L F-araA 39% ± 8%; 50 μmol/L araC 46% ± 5%. Evidently, araC is more toxic than araA or F-araA, likely owing to its more than 50-fold stronger inhibition of DNA replication (Supplementary Fig. S2) and is therefore used at 10-fold lower concentrations. The results in Fig. 1C indicate that while all nucleoside analogues exert marked radiosensitization (Supplementary Table S2), the effect of araA is superior in the sense that it combines large radiosensitization with acceptable toxicity. Therefore, in further mechanistic studies we focused on araA.

We inquired whether araA radiosensitization simply reflects its global effect on DNA replication. Therefore, we compared the radiosensitizing potential of araA with that of two widely used inhibitors of DNA replication, Aph and HU. As these compounds inhibit DNA replication with different efficiencies, we determined the IC₅₀ for each one (IC_50repl; Supplementary Fig. S2) and used it as orientation in the comparison of the radiosensitizing effect. The results summarized in Fig. 1D show that Aph at approximately 220-fold IC_50repl (6 μmol/L) fails to radiosensitize A549 cells. Also, HU at approximately 45-fold IC_50repl (2,000 μmol/L) causes only modest radiosensitization. On the other hand, araA at approximately 80-fold IC_50repl (500 μmol/L) causes marked radiosensitization. We conclude that effects more specific than global DNA replication inhibition underpin the radiosensitizing effect of araA. Indeed, araA is also a strong radiosensitizer of cells in phases of the cell cycle other than S-phase (31).

We inquired whether araA-mediated radiosensitization derives from inhibition of DSB repair. Actively growing, as well as 7-day old plateau phase A549 cells were irradiated and repair of DSBs was followed by PFGE (Fig. 2A and B). In exponentially growing A549 cells, treatment with araA at concentrations up to 2,000 μmol/L has no measurable effect on DSB repair up to 1 hour after ionizing radiation, and exerts only a small inhibitory effect after 2 hours that fails to reach statistical significance for all concentrations and time points, except for the 8 hour, 1,000 μmol/L time point (Fig. 2A). Similar treatment in plateau phase A549 cells fails entirely to generate detectable inhibition (Fig. 2B). On the other hand, treatment of exponentially growing A549 cells with the highly specific DNA-PKs inhibitor NU7441, markedly inhibits DSB repair as expected after specific inhibition of D-NHEJ (Fig. 2C).

This observation is surprising as previous experiments using CHO cells and neutral filter elution for DSB detection had shown a small but detectable inhibition of DSB repair by araA (2). Therefore, we analyzed DSB repair in irradiated highly enriched G₁- and G₂-phase populations sorted from araA-treated exponentially growing A549 cells as described previously (ref. 14; Fig. 3A). The results summarized in Fig. 3 indicate that FACS selects cell populations with excellent purity in G₁ and G₂, and that their distribution through the cell cycle remains practically unchanged during incubation for repair: 95.8% ± 2.0% for G₁ and 81.1% ± 6.0% for G₂ (Fig. 3D and E, right). Notably, both G₁ as well as G₂ cells repair DSBs effectively, and incubation with 1,000 μmol/L araA fails to produce detectable inhibition (Fig. 3D and E, left). The same holds true for S-phase–rich exponentially growing cell populations (Fig. 3C). We conclude that under the experimental conditions employed here, araA fails to inhibit global DSB repair in G₁, S, or G₂-phase of the cell cycle.

Extensive work carried out in our laboratory during the past 15 years suggests that experiments like those described above mainly reflect the function of D-NHEJ. This conclusion is supported by the observation that mutants deficient in HRR show unchanged potential for DSB repair by PFGE (32, 33). This is true even for cells in G₂-phase, where HRR unfolds its full potential (14). We therefore conclude that the above experiments conclusively demonstrate the unperturbed function of D-NHEJ in the presence of araA.

Analysis of phosphorylation of the histone variant H2AX (γH2AX) as an alternative to PFGE as endpoint to study the effect of araA on DSB repair was compromised by the pan-nuclear γH2AX staining observed specifically in the S-phase cell fraction (determined by EdU staining) already before ionizing radiation treatment (Supplementary Fig. S3). Similar effects have been reported for gemcitabine, HU, and thymidine and linked to inappropriate origin firing, as well as to replication fork stalling and collapse (34, 35). Therefore, alternative methods allowing specific analysis of the effects of araA on HRR or B-NHEJ were introduced.

We examined the effect of araA on HRR using RAD51 foci formation as endpoint (36). For experiments, actively growing A549 cells were pulse-labeled with EdU to mark S-phase cells. Cells were subsequently exposed to 4 Gy X-rays, incubated for 3 hours in the presence of different concentrations of araA, stained, and processed for analysis.

EdU-positive cells were considered as being in S-phase at the time of irradiation and were analyzed for RAD51 foci formation either during S- or in the subsequent G₂-phase. Cyclin B1–positive, but EdU-negative, cells were scored as G₂ at the time of irradiation and in the subsequent repair time interval. Cyclin B1 and EdU double negative cells were identified as G₁ at the time of irradiation and in the subsequent repair time interval, but were not analyzed in the context of the present set of experiments, as they fail to develop RAD51 foci after ionizing radiation (refs. 37, 38; Fig. 4A). In quantitative analysis, results from S- and G2-phase cells were pooled together, as the low fraction of G2-phase cells compromised their specific analysis in this experiment.

Figure 4B shows that upon exposure to ionizing radiation, RAD51 foci form promptly in S- and G₂-phase, as expected when HRR is functional. Notably, a marked reduction in the number of RAD51 foci is observed after treatment with araA, with a half maximum inhibition at about 250 μmol/L; RAD51 foci formation is nearly eliminated above 500 μmol/L araA. The reduction in RAD51 foci formation with increasing araA concentration is similar to that observed for radiosensitization at 3 Gy suggesting a cause-effect relationship between the two endpoints (Fig. 4C). We conclude that while araA has no effect on DSB repair by D-NHEJ, it exerts an impressively strong inhibitory effect on DSB repair by HRR, which correlates with the observed radiosensitization to killing.

We introduced functional assays to independently confirm the above presented effect of araA on HRR. For functional analysis of HRR, we used the human osteosarcoma cell line U2OS-282C bearing in its genome the DR-GFP reporter construct (28). In these cells, a DSB is generated within the reporter sequences by transfection and transient expression of I-SceI restriction endonuclease from the pCMV-3xNLS-ISceI plasmid, and its successful processing by HRR is reported by expression of GFP. A distinct characteristic of this assay is that GFP expression confirms the successful completion of all steps required for HRR. In good agreement with RAD51 immunofluorescence analysis, treatment with araA causes a strong inhibition of HRR in this functional assay as well (Fig. 5B).

We also tested the effect of F-araA and araC as well as of DNA replication inhibitors HU and Aph (Fig. 5B and C). It is evident that HU and Aph only marginally affect HRR. F-araA, on the other hand, exerts on HRR an effect very similar to that of araA, whereas araC shows a smaller inhibitory effect (Fig. 5C). We conclude that inhibition of HRR is a rather general property of nucleoside analogues, and that the magnitude of this inhibition can vary considerably among analogues.

Importantly, very similar trends are observed in experiments carried out with CHO cells carrying the DR-GFP construct (Supplementary Fig. S4). Thus, the strong inhibitory effect of araA on HRR is not species specific. CHO cells also served as a model to standardize the protocol of araA treatment. Supplementary Figure S4D shows that application of the 4-hour araA treatment window shortly after transfection gives maximum inhibition of HRR, whereas administration at later times reduces the effect.

Perhaps not unexpectedly, the effect of araA is transient on this HRR assay and reduced when cells are analyzed 48 hours after transfection (Supplementary Fig. S5). Three phenomena likely contribute to this effect: First, incubation without araA after the 4-hour treatment allows effective drug clearance and recovery of HRR. Second, sustained expression of I-SceI (for at least 2–3 days) enables DSB-induction long after removal of araA. Third, araA treatment-mediated enrichment of cells in S/G₂-phase (see Supplementary Fig. S5) promotes HRR. We conclude that araA strongly inhibits overall HRR function.

The strong, specific effect of araA on HRR predicts that the associated radiosensitization will be diminished in cells with impaired HRR. To test this hypothesis, we compromised HRR in A549 cells through RAD51 knockdown and tested araA radiosensitization. Western blotting shows successful reduction in RAD51 expression to less than 10% of the control (Fig. 5D). As reported earlier (39–43), reduced expression of RAD51 is associated with increased radiosensitivity to killing. Strikingly, while control cells show strong radiosensitization after treatment with araA, treatment with araA of cells treated with siRNA targeting RAD51 fail to show marked radiosensitization (Fig. 5D). The small radiosensitization observed at 5 Gy likely reflects the response of the less efficiently transfected cells, which remain partly proficient for HRR, and thus radioresistant and responsive to araA.

The lower cell plating efficiency (20% ± 6% as compared with 72% ± 6% in the controls) observed in cell populations treated with siRNA-targeting RAD51 is commonly found in cells with HRR defects and is, at least partially, a reflection of the efficient knockdown achieved. We conclude that in cycling cells, HRR proficiency is a prerequisite for efficient radiosensitization by araA.

Treatment of irradiated cells with araA has been shown to increase the frequency of exchange type chromatid aberrations (44), which suggests increase in the error-prone processing of DSBs. To specifically address this putative mode of araA action, we employed a U2OS cell line harboring a reporter construct (EJ5-GFP; Fig. 6A; top) that generates GFP signal only if the apical ends of two I-SceI–induced DSBs are rejoined by deletion of the intervening fragment, which by definition is a mutagenic event (28). Cells harboring this construct, when transfected with the I-SceI expression plasmid and analyzed 24 hours later, show a modest but robust increase in signal intensity after incubation with araA (Fig. 6A; bottom).

As the spectrum of errors reported by the EJ5-GFP substrate is limited to a single event, we introduced an alternative system designed to report a wide spectrum of deletions and other mutagenic events (29). In this two component system (Fig. 6B), mutagenic repair of an I-SceI–generated DSB inactivates the TetR gene, allowing thus expression of DsRED that can be detected by flow cytometry. Strikingly, in this assay, treatment with araA enhances mutagenic end joining two- to 3-fold (Fig. 6B; bottom).

SSA is also a mutagenic pathway of DSB repair. U2OS-283C cells carry a stable integration of the SA-GFP construct, specifically designed to report SSA events initiated by an I-SceI–induced DSB (Fig. 6C; bottom). As we have shown previously, knockdown of RAD51 or BRCA2 induces a strong increase in SSA due to suppression of HRR (23). Figure 6C shows that after treatment of U2OS-283C cells with araA, SSA is first suppressed as efficiently as HRR, but recovers to control levels at higher concentrations.

We considered that the araA-mediated inhibition of HRR alters the balance between HRR and SSA and causes the fluctuations in SSA noted in Fig. 6C. We measured therefore SSA after inhibiting HRR through knockdown of RAD51 (Supplementary Fig. S6). It is evident that following RAD51 depletion, araA exerts a monotonous albeit modest inhibition on SSA (Supplementary Fig. S6B). Thus, three independent reporter assays demonstrate that treatment with araA enhances error-prone DSB processing by either directly promoting error-prone repair processes, or by compensating for the loss of error-free repair pathways.

While the radiosensitizing potential of nucleoside analogues is well documented and extensively investigated, the underlying molecular mechanisms remain uncharacterized. Previous work suggested modest inhibitory effects of nucleoside analogue on global DNA repair (4) and more recent reports suggest that nucleoside analogues interfere with proper execution of HRR (45). Recent advances in our understanding of DSB repair and the discovery of alternative pathways of end-joining with backup function, generate an entirely new intellectual framework for the mechanistic analysis of radiosensitizing treatments, as shown for hyperthermia (46). Indeed, radiosensitization may be caused not only by selective inhibition of a specific DSB repair pathway, but also by a shift in the balance between them. Here, we attempt a mechanistic analysis of nucleoside analogue radiosensitization based on this premise, while focusing on araA.

The first novel and quite unexpected observation of our work is that D-NHEJ remains unperturbed by araA. This observation contrasts the strong radiosensitizing potential of araA and suggests that inhibition of other DSB repair pathways must play a defining role. Indeed a strong inhibition by araA of HRR is documented by scoring RAD51 foci formation, or by using specialized functional assays. Inhibition of HRR by araA correlates well with radiosensitization to killing suggesting cause-effect relationships. Vice-versa, cells radiosensitized by compromising HRR (Fig. 5D), show markedly reduced radiosensitization after treatment with araA.

The molecular mechanism of HRR inhibition by araA remains to be elucidated. AraA may inhibit the DNA synthesis step that follows strand invasion during HRR processing. However, the suppression of RAD51 foci formation is also compatible with interference with an earlier step of HRR. Collectively, these results identify HRR as a key target of nucleoside analogue radiosensitization.

Perhaps, the most unexpected and arguably one of the most far reaching observations of the present work is the observed promotion by araA of mutagenic DSB repair. This is striking particularly in view of the fact that D-NHEJ remains unaffected by araA.

Recent results from our laboratory show that B-NHEJ provides important backup functions not only for failed D-NHEJ but also for failed HRR causing translocations (11). This observation offers a rationale to explain the increase in mutagenic DSB repair observed after treatment with araA: By strongly inhibiting HRR, araA provides substrate for B-NHEJ, increasing thus its relative contribution to overall DSB processing, and consequently also the frequency of processing errors. Indeed, Mozdarani and Bryant reported an increase in translocations in irradiated cells after treatment with araA, which was unexpectedly paralleled by a global inhibition in the repair of chromosomal breaks (44).

SSA is also a homology-dependent, error-prone DSB repair pathway that is only mildly inhibited by low concentrations of araA, while benefitting from HRR inhibition at higher concentrations. The mutually exclusive interactions between SSA and HRR provide further evidence for the potential of araA to shift the balance between DSB repair pathways by inhibiting HRR (see Fig. 6).

There is a spectrum of nucleoside analogues with clinical application in the management of human cancer. The question arises, therefore, as to whether they all function via similar mechanisms as araA, and whether their mechanism of action as radiosensitizers can be predicted from other properties, for example, their ability to inhibit DNA replication.

The results presented here demonstrate differential radiosensitizing potential and different effects on HRR and B-NHEJ that may not be predicted from the ability of a nucleoside analogue to inhibit DNA replication. Such differences among nucleoside analogues in their ability to inhibit specific DSB repair pathways appears surprising considering their similar structures and accepted modes of action. Evidently, neither their efficacy in inhibiting replication nor their cytotoxicity are predictive of their efficacy as radiosensitizers.

Thus, interactions and interferences with the function of the HRR machinery, distinct from those occurring during DNA replication must be considered. These may involve inhibition of one or more of the participating enzymes in ways that cause abrogation of this form of DSB processing. In this regard, it is striking that small structural modifications are associated with marked differences in the ability of an analogue to inhibit a putative target protein, as evidenced by the resistance of F-araA to adenosine deaminase. Similarly, structural modifications may influence or even determine the differential effects of nucleoside analogues on DSB repair and cell radiosensitivity to killing.

Only after identification and characterization of the relevant target proteins will it be possible to rationally design novel compounds with improved activity. Our work makes an essential step forward in this direction by identifying HRR as the main, if not the only, repair pathway inhibited by nucleoside analogues, and its proteins as the first, natural, key candidate-targets. Indeed, all nucleoside analogues tested in the present study inhibit HRR, albeit to a different extent.

Furthermore, the question arises as to whether the correlation between inhibition of HRR and radiosensitization is universal. The results presented here suggest that this may not be the case. For example, araC is a stronger radiosensitizer than one would predict from its ability to inhibit HRR. Similarly, F-araA is slightly more effective than araA in inhibiting HRR in the reporter assay, but shows less radiosensitization than araA. Thus, additional mechanisms are likely to contribute to nucleoside analogue radiosensitization, and their elucidation will enhance the potential for clinical exploitation.

Central goal of the experiments described here was to obtain reliable mechanistic information on the radiosensitizing potential and HRR inhibition of the tested nucleoside analogues. At times, this led to their administration at concentrations higher than those typically achieved in the clinical setting. However, in the clinical setting factors like hematotoxicity, or the rate of metabolic activation/inactivation are key determinants of the drug concentrations that are, or can be achieved. These side effects may derive from molecular properties of the compounds that are independent from those causing radiosensitization. It is our working hypothesis that by investigating potentially unrelated effects separately, we generate information with direct utility not only in the optimization of the clinical administration of existing compounds, but also in the development of the next generation of nucleoside analogues.

Specifically, in the experiments described above araC was used at concentrations that can be clinically achieved in patients. For araA, concentrations below 100 μmol/L have been reported for patients (47, 48), which is at the beginning of the range used in our experiments. For F-araA concentrations between 2 μmol/L and 5 μmol/L (49, 50) have been reported which are much lower than those explored here.

We anticipate that once the range of concentrations is defined, where an effect can be observed, it will be possible to rationally design clinical administration protocols optimally exploiting this effect. Even when achievable drug concentrations are expected to produce small effects in single administration protocols, strategies can be devised, which amplify this small effect by appropriate integration in fractionated radiotherapy. Further improvements are possible by developing alternative delivery strategies maximizing the desired effect.

Our study provides evidence that a shift in the balance between error-free and error-prone DSB repair processes, without detectable changes in the overall efficiency of DSB rejoining by D-NHEJ underpins radiosensitization by nucleoside analogues. We report strong inhibition of HRR by the nucleoside analogues araA and F-araA and show that this inhibition contributes to radiosensitization. It will now be important to characterize the molecular mechanisms underpinning HRR inhibition by nucleoside analogues, as this is likely to pave the way for the development of compounds with optimized spectrum of activities.

No potential conflicts of interest were disclosed.

Conception and design: S. Magin, M. Papaioannou, J. Saha, G.E. Iliakis

Development of methodology: S. Magin, J. Saha, G.E. Iliakis

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Magin, J. Saha, C. Staudt

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Magin, M. Papaioannou, C. Staudt, G.E. Iliakis

Writing, review, and/or revision of the manuscript: S. Magin, M. Papaioannou, C. Staudt, G.E. Iliakis

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G.E. Iliakis

Study supervision: G.E. Iliakis

This work was funded by a grant (02NUK005C; awarded to G.E. Iliakis) by the German Federal Ministry of Education and Research (BMBF).

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.