Radiation Drives the Evolution of Orthotopic Xenografts Initiated from Glioblastoma Stem–like Cells

Glioblastoma (GBM) is the most common form of malignant brain cancer; despite the combination of surgery, radiotherapy, and temozolomide, the median survival time is only about 14 months (1). Although the mechanisms mediating this consistent therapeutic resistance have not been defined, the clonal diversity and evolutionary dynamics inherent to GBM are considered a major obstacle in the development of effective treatment (2–6). Along these lines, comparison of genomic data generated from glioma tissue obtained at initial surgery and at recurrence revealed an altered mutational profile, an effect that was attributed to temozolomide treatment (7). The implication of such studies is that the temozolomide-driven evolution results in the emergence of resistant clones. Consistent with studies of clinical specimens, temozolomide treatment of mice bearing brain tumor xenografts initiated from GBM primary cultures suggested the expansion of drug-resistant clones (8). Given that GBMs regrow after initial treatment, understanding the consequences of treatment-driven evolution may not only generate insight into the fundamental biology of recurrent GBMs but also suggest novel therapeutic strategies.

Although studies to date have focused on temozolomide (7, 9, 10), a role for radiotherapy as an independent driver of GBM evolution has not been investigated. Toward this end, orthotopic xenografts initiated from CD133⁺ GBM stem–like cells (GSC) would appear to provide a model system for testing the potential of radiation to influence GBM evolution. GSCs represent a clonogenic subpopulation considered to be critical in the development, maintenance, and treatment response of GBMs (11–13). Moreover, orthotopic xenografts grown from GSCs replicate the genotype, phenotype, and in vivo growth pattern of GBM (14). With respect to GBM evolution, we have previously shown that after the initial implant of 100% CD133⁺ cells, xenografts at the time of morbidity are comprised of a variety of cell subpopulations including those expressing GFAP or βIII tubulin (15), which is consistent with tumor cells that have differentiated, at least partially, along astrocytic and neuronal pathways, respectively. In addition, there continued to be a small subpopulation (∼10%) of tumor cells expressing CD133, suggesting the presence of GSCs. Finally, based on analysis of γH2AX and 53BP1 foci, CD133⁺ cells were less radiosensitive than CD133⁻ tumor cells (16). Thus, the GSC xenograft model exhibits the intratumor heterogeneity (ITH) and evolutionary dynamics that may simulate that of a GBM in situ.

To investigate the potential of radiotherapy to influence GBM evolution, in the study described here we defined the consequences of a fractionated radiation protocol on the growth pattern, clonal diversity, and genomic architecture of GSC-initiated orthotopic xenografts. The data presented show that tumors that regrow after irradiation were less invasive and had different mutational signatures as compared with untreated tumors. In addition, based on viral integration site analysis (VISA), radiation exposure resulted in a reduction in ITH (clonal diversity), an effect that was dependent on the brain microenvironment. These results indicate that radiation drives the evolution of the GSC-initiated orthotopic xenografts.

GSC lines NSC11 and NSC20 (provided by Dr. Frederick Lang, MD Anderson Cancer Center, Houston, TX, in 2008, as frozen stocks) were grown as neurospheres in stem cell medium and CD133⁺ GSC cells were isolated by FACS as reported previously (17). The U251 human GBM cell line was obtained from the Division of Cancer Treatment and Diagnosis Tumor Repository, NCI, and grown in DMEM supplemented with 10% FBS (Invitrogen). All cell lines were cultured less than 2 months after resuscitation; each tested negative for Mycoplasma by PCR. U251 cells were authenticated in July 2019 by STR Analysis (Idexx BioAnalytics); GSCs were authenticated by routine morphologic and growth analysis. All lines were transduced with lentivirus (LVpFUGQ-UbC-ffLuc2-eGFP2) at a multiplicity of infection of 1 (16). For in vitro experiments, GSC neurospheres were disaggregated into single cells (17) and seeded onto poly-l-ornithine (Invitrogen)/laminin (Sigma-Aldrich)-coated tissue culture dishes in stem cell media. Under these conditions, single-cell GSCs attach and grow as an adherent monolayer maintaining CD133 expression and stem-like characteristics (18). Radiation was delivered in vitro using a 320 kV X-ray machine (Precision X-Ray Inc.; ref. 19).

CD133⁺ GSCs (1.0 × 10⁵) or U251 (2.5 × 10⁵) transduced to express luciferase and GFP were implanted into the right striatum of 6-week-old athymic female nude mice (NCr nu/nu; NCI Animal Production Program; ref. 16). Bioluminescent imaging (BLI) and local irradiation were all performed as described previously (16). On day 6 (U251) or day 21 (NSC11 and NSC20) after implantation, mice were randomized according to BLI signal into two groups: control and irradiated (RT). For GSC tumors, 5 Gy was delivered for 3 consecutive days; for U251 tumors, 3 Gy was delivered on 3 consecutive days. Radiation was delivered using an X-Rad 320 X-irradiator (Precision X-Rays, Inc.), dose rate of 2.9 Gy/minute. Mice were monitored daily until the onset of neurologic symptoms (morbidity). BLI was performed weekly after irradiation until the first mouse of the group was lost. GraphPad Prism 7 (GraphPad Software) was used to generate Kaplan–Meier survival curves. To generate leg tumors, 1.0 × 10⁶ U251 cells were subcutaneously injected into the right hind leg. When tumors were 150 to 200 mm³, mice were randomized; radiation was delivered locally with animals restrained in a custom designed jig. Subcutaneous tumors were collected for analysis at 1,500 mm³ or 2 cm in any dimension. All experiments were performed as approved by the NIH Guide for Care and Use of Animals and conducted in accordance with the Institutional Animal Care and Use Committee.

At the initial signs of morbidity, mice were euthanized and perfused with chilled PBS then formalin via cardiac puncture. Brains were removed, placed in 10% buffered formalin before sectioning, embedded in paraffin, and cut into 6-μm-thick slices. After heat-induced epitope retrieval and blocking slides were incubated with SOX2 primary antibody (anti-human; Cell Signaling Technology) followed by horseradish peroxidase secondary (anti-rabbit; Vector Laboratories). Slides were exposed to DAB Reagent (Vector Laboratories), counterstained with hematoxylin, and visualized with an AxioScan Imager (Zeiss).

VISA was performed by the Center for Cancer Research Genomics Technology Laboratory (20, 21). Briefly, genomic DNA was sheared to an average size of 400 bp and subjected to linker-mediated, nested PCR using a combination of long terminal repeat (lentivirus) and linker-specific primers. Illumina sequencing adaptors were added at the same time. The library was sequenced on Illumina MiSeq using 2 × 150 bp PE reads. Integration site junctions were mapped to hg19 human reference genome. Insertion sites are expressed as a percentage of the total reads. To compare samples within an experiment, integration sites detected in at least two of the samples were subjected to unsupervised hierarchical cluster analysis using R to visualize relative changes in clonal diversity.

Genomic DNA was subjected to whole-exome sequencing (WES), which was performed by the Center for Cancer Research Genomics Technology Laboratory. Extracted DNA underwent library prep according to the Agilent SureSelect XT (All Exon V5 +UTR) protocol and sequenced on Illumina HiSeq4000 device using paired-end sequencing to an average sequencing depth of >180×. Mouse reads were removed as described previously (22). Alignment and tumor-only variant calling was performed with the Center for Cancer Research Collaborative Bioinformatics Resource pipeline (https://github.com/CCBR/Pipeliner). Germline variants were excluded using a panel of normals (23). Only protein-altering variants were retained (24). COSMIC mutational signature analysis (https://cancer.sanger.ac.uk/cosmic/signatures_v2) was performed with the YAPSA R package (v1.8.0; http://bioconductor.org/packages/release/bioc/vignettes/YAPSA/inst/doc/YAPSA.html) as described previously (25, 26). Tumor subpopulations were further defined using EXPANDS algorithm and R package (v2.1.2) with default parameters (27). EXPANDS determines cell-frequency probabilities by using copy-number variation (CNV) and single-nucleotide variants (SNV) allele frequencies to estimate the fraction of cells harboring a SNV (SNVs and CNVs are filtered to remove variants on the X and Y chromosomes; ref. 27). For COSMIC and EXPANDS analyses, variants were included when they were protein-altering, present in at least two of three replicates, and had variant allele frequency (VAF) > 5%. Sequence data have been deposited in NCBI's BioProject database and are accessible through ID PRJNA576782 (http://www.ncbi.nlm.nih.gov/bioproject/576782).

As a model system for investigating the impact of radiotherapy on GBM evolution, we used orthotopic xenografts initiated from CD133⁺ GSC lines. GSCs were transduced with a lentivirus containing a bimodal expression vector fused with the bioluminescent protein ffLuc2 and fluorescent protein eGFP2 under UbC promoter control (LVpFUGQ-UbC-ffLuc2-eGFP2; ref. 16). After in vitro expansion, cells were implanted into the right striatum of nude mice. At 21 days postimplant, upon reaching a size consistently detectable by BLI, tumor-bearing mice were randomized into treatment groups, control (mock), and 3 × 5 Gy. As shown in Fig. 1Aand B, this fractionated irradiation protocol resulted in significant survival increases for NSC11 or NSC20 tumor-bearing mice. This survival advantage was consistent with the delay in growth rate reflected by BLI as a function of time after irradiation (Fig. 1C and D). Given that 3 × 5 Gy initiated at day 21 postimplant significantly delays tumor growth yet does not achieve curative effects, a situation not unlike that typically observed in clinic, we used this treatment protocol to test the hypothesis that radiation drives the evolution of GSC-initiated brain tumor xenografts.

As an initial assessment of the potential for radiation to influence GBM evolution, the growth patterns of control and irradiated GSC xenografts were compared. GSCs engineered to express GFP can be visualized in tumor-bearing brains using a stereomicroscope. Representative images (Fig. 2A) of control mice demonstrate GFP signal from NSC11 or NSC20 cells extends anteriorly and posteriorly throughout most of the right hemisphere, well beyond the right striatum implantation site, reflecting a highly diffuse tumor (corresponding bright field images are shown in Fig. 2B). However, in irradiated mice, GFP fluorescence was more restricted and limited primarily to the anterior portion of the hemisphere. To determine whether these observations extended to the histologic level, right hemisphere sagittal sections from control and irradiated NSC11- and NSC20-bearing brains were evaluated for human SOX2 staining, which is highly expressed in both lines. As shown in Fig. 2C; Supplementary Fig. S1, GSCs in control mice were diffusely distributed with poorly defined margins and gray and white matter infiltration. In mice that received fractionated radiation, tumors were less infiltrative with more clearly demarcated borders as compared with controls. Although indicative of an altered growth pattern, in broader terms these results suggest that the fundamental biology of GSC-initiated tumors that regrow after irradiation diverges from that of untreated tumors. As a possible explanation, we hypothesized that morphologic/histologic modifications detected in recurrent xenografts reflected the radiation-induced selection of tumor cell subpopulations, that is, radiation-driven evolution.

To directly address the question as to whether radiation modifies the clonal diversity (i.e., intratumoral heterogeneity) of GSC-initiated xenografts, VISA was used (20). Facilitating the application of VISA, our in vivo studies have centered on GSCs transduced with the lentivirus LVpFUGQ-UbC-ffLuc2-eGFP2, which, as discussed above, allows for in vivo tumor growth monitoring and ex vivo visualization. In addition, the stable integration of the lentivirus into the genome at essentially random sites provides a unique tag for each cell in the transduced population, which is inherited by subsequent daughter cells. Because the integration site is largely within a transcriptional unit, sequencing of PCR-amplified genomic DNA surrounding the lentiviral sequence allows for assigning each site to a single location/gene or nearest gene. Identifying the gene thus merely provides a method for easily cataloging the integration site. The percent reads for a given gene/site corresponds to the size of the clone within the transduced population.

For this experiment, CD133⁺ NSC11 cells were transduced with the described lentivirus; the culture was expanded and used for orthotopic implantation or collected for in vitro VISA. At 21 days postimplant, as for the survival studies, mice were randomized according to BLI into two groups: control (ic) and irradiated (icRT, 3 × 5 Gy). At morbidity, mice were euthanized, GFP-expressing brain tissue grossly dissected, and DNA was extracted for VISA. The percent reads detected at a given integration site (clonal frequency) were visualized for the individual xenografts and in vitro cultures in the heatmap shown in Fig. 3A. As compared with NSC11 cells grown in vitro, there was a reduction in the number of clones detected (unique integration sites) in each of the four untreated xenografts (ic), suggesting a selection for clones that preferentially grow under in vivo orthotopic conditions. Moreover, as compared with the untreated tumors (ic), there was a further reduction in the total number of individual clones detected in the four xenografts that had been irradiated (icRT).

VISA and unsupervised hierarchical cluster analysis of clonal frequency were also performed on NSC20 samples (Fig. 3B). Although there was variability between xenografts in the control (ic) and irradiated (icRT) groups, as compared with NSC20 in vitro, there was a reduction in clonal diversity in each of the orthotopic xenografts, with a further reduction evident in the irradiated tumors, similar to NSC11. The VISA results were also expressed in Fig. 3C and D, as the average number of unique integration sites detected in each group. As shown, the largest number of integration sites was detected under in vitro conditions, which was reduced in orthotopic xenografts and further reduced after irradiation of xenografts. These results suggest that only a subset of the NSC11 and NSC20 cells proliferating in vitro will form a tumor after orthotopic implantation. Moreover, of those clones that grow orthotopically, after irradiation there is an additional reduction in clonal diversity, which is consistent with radiation-driven tumor evolution.

To investigate the role of the microenvironment in the radiation-induced reduction in clonal diversity, VISA was performed on NSC11 cells irradiated in vitro. Twenty-four hours after plating, cultures received either mock (control), 4 Gy (single dose), or 3 × 2 Gy. Cells were harvested for VISA when they reached 70% to 80% confluency, which corresponded to 7, 19, and 21 days for control, 4 Gy, and 3 × 2 Gy, respectively. Clonogenic survival analysis showed that the surviving fraction of NSC11 cells exposed to 4 Gy and 3 × 2 Gy was 0.008 ± 0.004 and 0.039 ± 0.031 (mean ± SEM, n = 3), respectively. As shown in Fig. 4A, cultures did not cluster according to treatment group. Moreover, there was no significant difference in the average number of integration sites among the three groups (Fig. 4B). These results suggest that, in contrast to the orthotopic model, in vitro irradiation had no detectable effect on clonal diversity. These data suggest that the radiation-induced reduction in clonal diversity is dependent on the in vivo microenvironment.

To determine whether radiation-driven evolution was limited to xenografts initiated from GSCs and to further interrogate the role of the brain microenvironment, VISA was applied to the established glioma cell line U251. In this experiment, lentivirus transduced U251 cells were grown as intracerebral (ic) xenografts and as subcutaneous leg tumors. At 6 days after ic (when tumors were detectable by BLI) and 18 days after subcutaneous implantation (150–200 mm³), U251 tumors received 3 Gy for 3 consecutive days (3 × 3 Gy). Irradiated and control ic tumors were collected at morbidity and subcutaneous tumors when the volume exceeded 1,500 mm³. Unsupervised hierarchical cluster analysis of the clonal frequency of the tumor samples as compared with U251 grown in vitro is shown in Fig. 4C; the average number of integration sites in each group is shown in Fig. 4D. Although there was no consistent difference between control ic and subcutaneous tumors, the number of integration sites were reduced in each compared with U251 in vitro, although not to the degree as observed for the GSC ic tumors. In response to irradiation, the number of integration sites in subcutaneous tumors is reduced, suggesting a decrease in clonal diversity. However, in ic U251 tumors, irradiation induced a greater reduction in integration sites, suggesting that the radiation-mediated reduction in clonal diversity is not limited to xenografts initiated from GSCs. Moreover, these results suggest that the brain microenvironment plays a critical role in radiation-driven evolution of GBM xenografts.

Because investigations of tumor evolution typically apply WES to describe genetic ITH, we used this approach to evaluate the potential for radiation-driven evolution of GSC-initiated orthotopic xenografts. Genomic DNA samples extracted from morbid tumors treated as in the VISA experiments shown in Fig. 3 were utilized for WES. Over 5,500 SNVs + indels were detected in NSC11 tumors, of which, the majority were shared between control and irradiated tumors (Fig. 5A). Variants unique to control or irradiated tumors (private mutations) were consistently detected among the three samples in each group (Fig. 5B). The shared and private mutations in control and irradiated tumors were classified according to the mutational signatures generated from patient characteristics within large cancer cohorts using COSMIC mutational signature profiling (25, 28, 29). As shown in Fig. 5C, the shared mutations distributed to mutational signatures 1 (deamination of 5-methylcytosine), 5 (unspecified etiology), 11 (temozolomide treatment), 20 (concurrent POLD1 mutation and mismatch repair deficiency), and 26 (defective DNA mismatch repair). Mutations unique to control tumors associated with similar signature profiles. However, private mutations found in irradiated tumors did not include signatures 1, 5, and 20, but mapped to signatures 3 (defective HR DNA repair: BRCA1/2 mutation), 6 (defective DNA mismatch repair), and 15 (defective DNA mismatch repair), which were not associated with the unirradiated tumors. These results suggest the emergence of different subpopulations in irradiated samples as compared with control NSC11 tumors.

These analyses were also performed on WES data generated from control and irradiated NSC20 tumors (Fig. 6). In NSC20 tumors over 6,000 SNVs + indels were detected with most of the variants shared between control and irradiated tumors with a similar number of private mutations detected only in control or irradiated NSC20 tumors (Fig. 6A and B). Using the shared mutations between control and irradiated NSC20 tumors, the mutational signatures identified were similar to those in NSC11 tumors (signatures 1, 5, 11, 20, and 26; Fig. 6C). In contrast, the signatures defined using the private mutations found in irradiated NSC20 tumors corresponded to increases in signatures 5 (unspecified etiology) and 19 (unspecified etiology) and decrease in signature 11 (temozolomide treatment). Although these signatures were substantially different from those identified for irradiated NSC11 tumors, as for NSC11, these analyses suggest that irradiation modified the subpopulations comprising NSC20 tumors, consistent with radiation-driven evolution.

To investigate the effects of radiation on xenograft subpopulation heterogeneity in more detail, the bioinformatics tool EXPANDS was applied to the WES dataset. EXPANDS utilizes VAFs and copy-number data to estimate cell-frequency probabilities (27). For this analysis, variants that were detected in at least two of the three replicates per group (control and RT) were included. Figure 7A depicts the 10 predicted subpopulations found in control and irradiated NSC11 tumors with the largest subpopulation (blue) at the bottom of the graph; each black dot corresponds to a variant predicted to be in a subpopulation. The lines indicate whether a specific variant in control tumors appeared in a subpopulation of the same ranking in the irradiated tumors or whether it had moved to another ranking. Although most variants are conserved between the largest subpopulations of control and irradiated tumors (depicted by straight lines), there was also a considerable amount of shifting from control tumors to a subpopulation of a different prevalence in irradiated tumors, which is especially apparent in comparisons of the smaller subpopulations (Fig. 7A, top). Figure 7B shows 13 predicted subpopulations for NSC20 tumors. Again, the largest, most prevalent subpopulation (blue) is at the bottom of the graph. Similar to NSC11, although the majority of variants are conserved within the largest subpopulations, there is still substantial subpopulation shifting for variants between control and irradiated tumors.

These analyses were extended to driver gene mutations previously associated with GBM (30–33). Initially, as shown in Fig. 7C and D, variants in driver genes were identified in each of three control and three irradiated GSC xenografts and classified as clonal (present in largest subpopulation), subclonal (present in smaller subpopulation), or both (multiple mutations for a given gene were individually found in both clonal and subclonal populations). In NSC11 tumors (Fig. 7C), multiple driver gene mutations were consistently lost in irradiated tumors, including AKAP9, FN1, SOX2, and SPTA1. There were also mutations that shifted in clonal status from control to irradiated tumors such as MAP4K3. Although there was more variability between individual tumors within control and irradiated groups, irradiation also resulted in changes in driver gene mutations detected in NSC20 tumors (Fig. 7D). Lost consistently after irradiation of NSC20 tumors were mutations in PIK3CA and CARM1. These data suggest that irradiation influenced the presence and clonality of GBM driver mutations.

The shifts of these GBM driver gene mutations between the previously defined subpopulations were then assessed (only subpopulations containing a GBM driver gene variant were included). As in the overall WES data (Fig. 7A), the majority of NSC11 GBM driver gene mutations were conserved within the largest subpopulations in control and irradiated samples (Fig. 7E). However, there were several mutations that became less prevalent after irradiation (e.g., lines from control subpopulation 1 to RT subpopulations 2 and 4). There was also one mutation that become more prevalent after irradiation (line from control subpopulation 5 to RT subpopulation 1), suggesting that this subclone may have had a better survival rate after radiotherapy. Figure 7F displays similar subpopulation dynamics for NSC20 tumors, although more variant shifts between subpopulations were noted for NSC20 compared with NSC11, including at least three variants that increased in prevalence after treatment. Changes in the distribution of mutations as a function of subpopulation size between control and irradiated tumors are consistent with subclone expansion and contraction, that is, subpopulation evolution. Taken together, WES data support the radiation-driven evolution of orthotopic xenografts initiated from GSCs.

The goal of this study was to test the hypothesis that radiation drives the evolution of orthotopic xenografts initiated from GSCs. Given that ITH is a critical parameter mediating tumor evolution, VISA was used to assess the changes in clonal diversity existing in cell lines and tumor xenografts. In these analyses, in the absence of any radiation exposure, comparison of the GSC lines grown in vitro and grown as brain tumors showed a significant reduction in the clonal diversity with intracerebral growth. These results imply that, although isolated by CD133 expression, the GSC cultures are comprised of a heterogeneous population, which is consistent with previous reports (34–37) and that only selected subpopulations were capable of proliferation under orthotopic conditions. As compared with untreated tumors, there was a dramatic additional reduction in clonal diversity in GSC-initiated tumors that had been irradiated. These results are suggestive of radiation-induced selection of subpopulations within GSC xenografts with the ultimate consequence of a reduction in ITH. The radiation-induced reduction in clonal diversity was also detected in orthotopic xenografts initiated from the long-established GBM cell line U251, indicating that the process is not unique to the GSC model and suggesting that radiation-driven evolution may be applicable to brain tumors in general.

Further analyses indicated that the reduction in ITH was dependent on the brain microenvironment. No change was detected in clonal diversity when VISA was performed on NSC11 cells irradiated in vitro. Moreover, comparison of U251 subcutaneous and ic xenografts showed that while there was a reduction in clonal diversity after irradiation of subcutaneous tumors, the effect was considerably more pronounced in ic U251 tumors, a reduction comparable to those detected in the GSC tumors. The implication of this reduction in ITH is that the brain microenvironment imposes the selection of subpopulations of GSC and U251 cells. Such a subpopulation selective effect could be attributed to normal tissue influence dependent on intrinsic characteristics of the cells implanted (e.g., genotype) or localization into a niche independent of tumor cell type. Although the mechanisms remain to be determined, the results presented emphasize the need to account for the brain microenvironment when investigating GBM radioresponse.

As an alternative approach to evaluating radiation-driven evolution, studies were extended to WES, which has been the standard approach to evaluating tumor evolution. Comparison of untreated GSC tumors and tumors that had regrown after irradiation revealed a consistently different set of mutations as well as mutational signatures. There also appeared to be a radiation induced shift in the genetic composition of the ranked subpopulations comprising the NSC11 and NSC20 xenografts. In evaluating the driver genes associated with GBM (30–33), there was a reduction in the number of mutations in irradiated NSC11 tumors, which would be consistent with a reduction in ITH. In NSC20 tumors, although there was also a reduction in the number GBM-associated gene mutations, the number of losses as well as the specific genes affected were different as compared with NSC11. The different mutation spectrums observed for NSC11 and NSC20 tumors with and without RT suggest that whereas radiation drives the evolutionary process, the specific events along with the functional consequences may be tumor dependent. Thus, whereas WES results are indicative of radiation-driven evolution, they do not provide any clear insight into the mechanisms involved or the characteristics of the surviving tumor cells. Of note, the reduction in mutations detected in the irradiated xenografts contrasts with the hypermutation reported for tumors treated with temozolomide (5, 7, 38, 39). Whether temozolomide, a component of GBM standard of care, influences radiation-driven evolution remains to be determined.

If radiation alters GBM evolution, then the fundamental biology of the tumors that recur after an initially effective course of radiotherapy should be altered. An example appears to be the morphology/histology of the NSC11 and NSC20 xenografts that regrow after the fractionated radiation protocol as compared with untreated tumors. As shown, each tumor exhibited a more restricted growth pattern in response to irradiation. Although in conflict with reports showing that radiation increases the invasion propensity of glioma cells in vitro (40, 41), these results agree with clinical observations indicating that >80% of GBM recur in the initial radiation treatment volume (42, 43). The studies described here show that radiation alone drives the evolution of GSC orthotopic xenografts affecting their ITH, mutation profile, and growth pattern. If a similar process is operative in a clinical setting, then therapeutic targets in a primary, untreated GBM may differ from those in the recurrent tumor, as previously suggested (44). GSC xenograft tumors that regrow after an initial radiation exposure may thus provide a model system for testing therapies for recurrent GBM. This would include re-irradiation alone as well as in combination with potential radiosensitizers.

No potential conflicts of interest were disclosed.

The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

Conception and design: J.H. McAbee, K. Camphausen, P.J. Tofilon

Development of methodology: B.H. Rath, X. Wu, P.J. Tofilon

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.H. McAbee, B.H. Rath, D.L. Young, X. Wu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.H. McAbee, B.H. Rath, K. Valdez, X. Wu, U.T. Shankavaram, K. Camphausen, P.J. Tofilon

Writing, review, and/or revision of the manuscript: J.H. McAbee, D.L. Young, X. Wu, K. Camphausen, P.J. Tofilon

Study supervision: P.J. Tofilon

Financial support for this study was provided by the Division of Basic Sciences, Intramural Program, NCI (Z1ABC011372) to P.J. Tofilon. This project has also been funded in whole or in part with federal funds from the NCI, NIH, under contract no. HHSN261200800001E.

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.