The Polyamine Catabolic Enzyme SAT1 Modulates Tumorigenesis and Radiation Response in GBM

Glioblastoma multiforme (GBM) is the most common and aggressive of the gliomas, a group of tumors that derive from glia or their precursors in the central nervous system (CNS). Patients with GBM have a median survival time of approximately 12 months, and only 3% to 5% of the patients survive more than 3 years (1). GBM is characterized by a heterogeneous population of cells that are genetically unstable and infiltrative, and comprise some of the most challenging therapeutic targets due to their anatomic location, the blood barrier, and poor responses to conventional therapies (2). Although the standard of care for newly diagnosed patients includes resection, followed by concurrent radiotherapy and temozolomide (3), the response to radiation and DNA-damaging agents remains insufficient as tumors display resistance and a propensity to recur. Although several mechanisms have been proposed to explain the radioresistance in GBM (4–6), the molecular bases of radioresistance remain incompletely defined.

To identify novel mediators of radiation resistance in glioblastoma, we performed a genetic screen using an shRNA library on two GBM cell lines. The results showed a list of overlapping genes with a variety of disparate functions. Analyses of the genes using public databases revealed spermidine/spermine-N1-acetyltranferase (SAT1) as a novel regulator of radiation response with no previously described association with DNA repair. SAT1 catalyzes the acetylation of polyamines spermidine and spermine to form acetyl derivatives and is considered a rate-limiting enzyme in polyamine catabolism, leading to degradation or excretion (7). Paradoxically, polyamines, ubiquitous cationic molecules, are known radioprotectors through their capacity to compact DNA (8–10), confounding speculation of how SAT1 could promote resistance to radiation.

Histone acetylation has been shown to have an essential role in DNA repair allowing critical proteins to be loaded at sites of damage (11), as well as altering gene expression by decompacting chromatin (12). Indeed, cells with DNA breaks maintain high levels of acetylation (13). In a parallel role to histone acetylation, we hypothesized that polyamine acetylation by SAT1 may have an integral role in double-strand break (DSB) repair through alteration of chromatin, and thereby contribute to radiation resistance. Our results indicate that SAT1 increases acetylation of histone H3, increasing BRCA1 expression, and allowing activation of the homologous recombination (HR) pathway to repair DNA damage. The findings support a novel role for SAT1 in histone acetylation and DNA repair, and suggest that the biologic significance of SAT1 expression in GBM lies in its contribution to radio- and chemoresistance. SAT1 may represent a therapeutic target to sensitize GBM to cancer therapies.

U87MG and LN229 cells were obtained from the ATCC. D54MG and D317MG lines were gifts of Dr. Jeremy Rich (Cleveland Clinic Foundation, Cleveland, OH); MCF7 DR-GFP cells were from Dr. Junran Zhang (Case Western Reserve University, Cleveland, OH). Neurosphere cell lines GBM 0821 and 0913 were a gift of Dr. Angelo Vescovi (University of Bicocca, Milan; to E.E. Bar). The lines were not independently authenticated. All lines were used in early passage. D-luciferin came from Byosynth International. Trichostatin A (TSA) was used at 250 ng/mL (Sigma-Aldrich). Antibodies: BRCA1 (sc-6954; Santa Cruz Biotechnology), Mre11 (cs4847; Cell Signaling Technology), NBS1 (cs3002), Rad51 (sc-8349), Histone H3 (sc-8654), acetyl-H3 (06-599; Upstate-Millipore), SAT1 (sc-67159). Quantitative real-time PCR (qRT-PCR) was performed using 2XSYBR-Green Master mix (Roche) and normalized to β-actin. Primer sequences: SAT1 F-5′GCTGATCAAGGAGCTGGCTA-3′, R-5′CAACAATGCTGTGTCCTTCC-3′; BRCA1 F-5′TGGAAGAAACCACCAAGGTC-3′, R-5′ACCACAGAAGCACCACACAG-3′; Actin F-5′CATGTACGTTGCTATCCAGGC-3′, R-5′CTCCTTAATGTCACGCACGAT3′; ODC F-5′ACATCCCAAAGCAAAGTTGG-3′, R-5′AGCTGACACCAACAACATCG 3′.

The Decode RNAi Screening Library was performed as directed (Open Biosystems). Microarray hybridizations were performed by the Stanford Microarray Facility. For siRNA, cells were transfected with DharmaFECT#1 and either 25 nmol/L control oligos or siSAT1 (Dharmacon). Stable SAT1 knockdown was performed with lentiviral shRNA pLKO.1 clones: TRCN0000035250 and TRCN0000035252 (Thermo Fisher Scientific), shRNA GFP was used as control. Cells were selected with puromycin at 1 μg/mL.

Radiation was performed with a ¹³⁷Cs irradiator (Shepherd). A total of 500 to 10,000 cells per plates were stained after 10 days with 0.1% crystal violet. Assays were done ≥3 times with individual samples in triplicate. Cell sorting was performed on live D54MG cells stained with Hoechst 33342 on an iCyt Reflection. Cells were then irradiated and plated. Clonogenic assays of neurosphere lines were performed in plates coated with 200 μL of 2 mg/mL poly (2-hydroxyethyl methacrylate; Sigma) in 95% ethanol. Wells were overlaid with 2 mL of neuro stem cell (NSM) 1.5% methyl-cellulose media. A total of 10⁴ cells in NSM were mixed 1:4 with methyl cellulose, and fed every 3 days. Sphere formation was monitored and scored by light microscopy after 12 days with Metamorph software, using a size cutoff of 100 μm.

Cells were fixed in 3% formaldehyde and stained using standard procedures: BRCA1 antibody (sc-6954), secondary Alexa-Fluor 594 anti-mouse (A11032; Invitrogen). Slides were mounted in Vectashield with DAPI (Vector Laboratories). Immunofluorescence was observed at ×100, foci were counted from at least 50 cells.

Eight-week-old athymic BALB/c mice were injected intacranially with 10⁵ luciferase-expressing SAT1 or GFP knockdown GBM0821, GBM0913 neurosphere cells. Tumor growth was monitored twice weekly and quantified using bioluminescent imaging. Signal intensity was measured as photon counts within a region of interest. Animal appearance, behavior, and weight were monitored to evaluate tumor progression. Subcutaneous U87MG tumors were produced by injection of 5 × 10⁶ cells in nude mice, and measured with calipers twice weekly. When tumors reached between 100 and 200 mm³, animals were randomized and injected intratumorally with 500 pmol/L of siRNA packaged in ECO (1-aminoethyl)iminobis[N-(oleicylcysteinyl-1-ami no-ethyl)-propionamide] as described previously (14). After 48 hours, half of the tumors were irradiated.

Chromatin immunoprecipitation (ChIP) was performed as described previously (15). The primers used were: BRCA1 F-5′GGCAGGCATTTATGGCAAAC-3′, R-5′TTCGGAAATCCACTCTCCCACG-3′. β-Actin F-5′CCGAAAGTTGCCTTTTATGG-3′, R-5′CAAAGGCGAGGCTCTGTG-3′. Acetylated H3 antibody (Upstate-Millipore 06-599). Samples were normalized to input.

Polyamines were quantified using mass spectrometry. Proteins were precipitated with heptafluorobutyric acid, and supernatants containing polyamines were filtered through an ion-exchange membrane for salt removal. An internal standard, diethylspermine, was added into the samples that were injected into an Agilent 6420 triple quadrupole LC/MS system for analysis along with a 7-point standard curve. The chromatographic separations were achieved using a guarded Luna reversed phase CN column (3 × 150 mm; Phenomenex). Concentrations were calculated from the standard curves. Three biologic replicates were assayed for each sample.

Cells were transfected with 2 μg of pcDNA or I-Sce plasmids using lipofectamine 2000 (Invitrogen), incubated 48 hours, and analyzed by flow cytometry as described previously (16). Comet assays were performed as directed (Trevingen), and quantified with ComeScore software. At least 41 cells per sample were measured, and done in duplicate.

Student t tests were used throughout the study to test the significance of differences between samples. Survival analyses were performed by log-rank tests in GraphPad Prism.

To identify novel mediators of radiation resistance in GBM, we performed a genetic screen using a lentiviral-mediated shRNA library on two GBM lines. U87MG and LN229 were chosen based on known different genetic characteristics to maximize the application of hits. U87MG is p53 wild-type and PTEN-null; LN229 is p53 mutant and PTEN wild-type (17, 18). Thus, many classic DNA damage pathway genes will be excluded because of the differential status of p53. The library was comprised of roughly 30,000 barcoded shRNAs, which target more than 10,000 genes. GBM cells were infected with pools of lentiviruses at an estimated multiplicity of infection of 0.3 to ensure that cells were infected with only once. Following selection, cells were divided into treatment (2 Gy of IR) and control groups, maintained for 72 hours after irradiation, and lysed for genomic DNA. The screens were performed in duplicate.

Ostensibly, shRNA knockdown of genes that confer radioprotection should result in the sensitization of cells to radiation and depletion from the population. Knockdown of genes that promote sensitivity to radiation should result in protection and increased survival. Identification of shRNAs that altered the sensitivity of cells to radiation was performed by determining the relative abundances of the shRNA barcodes in the genomic DNA of the treatment group versus the control group. Barcodes were amplified by PCR, labeled with Cy3 or Cy5, and hybridized onto barcode microarrays. The raw data of the U87MG screen are represented in Fig. 1A: Log₂-transformed ratios of barcode abundances in the irradiated sample divided by the control. Of 21,555 detected barcodes, 1,868 were decreased by 1.5-fold or more; 764 were decreased by 2-fold or more; and 126 were decreased by 4-fold or more (Fig. 1B). As an internal positive control for the screen, 47 shRNAs targeting genes known to be involved in DNA repair were found (e.g., ATM, PARP1, RAD9, and RAD51). When the results of the U87MG and LN229 screens were combined into one dataset, 79 shRNAs that were decreased 1.5-fold or more were identified, 10 were decreased 2-fold or more, and none were decreased by 4-fold. Genes with functions as broad ranging as synaptic nerve transmission to cell motility were included in the 10 genes (Supplementary Table S1; Microarray data are available in the ArrayExpress database, accession number E-MTAB-2861). Because of its metabolic role as a regulator of DNA compaction, we further investigated SAT1.

As a primary filter for relevance to GBM, expression levels of hits were queried in Oncomine. SAT1 expression was found elevated in a variety of brain and CNS cancers, derived from data from five studies of glioblastoma, two oligodendroglioma, one malignant glioma, one astrocytoma, and one oligoastrocytoma. The expression levels of SAT1 in cancers versus normal tissues in four of the studies are displayed in Fig. 1C. In the Shai, Bredel, and The Cancer Genome Atlas (TCGA) studies of GBM samples, SAT1 was overexpressed by factors of 3.11 (P < 0.0001), 3.33 (P = 0.0009), and 2.28 (P < 0.0001), respectively; in the Pomeroy study of malignant gliomas, SAT1 was overexpressed by 3.33 (P = 0.0042). We next looked at SAT1 Rembrandt, an NCI/NINDS database of gene expression and survival data from >340 brain tumor cases. Using SAT1 expression to categorize the samples, we observed that a 2-fold increase in SAT1 expression correlated with a significant decrease in survival time (603 vs. 450 days, P = 0.0008, log-rank test; Fig. 1D). Levels of SAT1 could not identify patients with poorer or better outcomes among patients with GBM, but could delineate patients with glioma with the poorest prognoses (i.e., patients with GBM) from the rest. Together, these data suggest SAT1 may play a role in GBM tumors and their radioresistant phenotype.

To verify SAT1 knockdown can sensitize cells to radiation, two unique lentiviral shRNAs were stably expressed first in U87MG cells, and clonogenic survival assays were performed with different doses of ionizing radiation (IR). qRT-PCR and Western blot analysis were used to verify knockdown (Fig. 2 insets; and Supplementary Fig. S1). Importantly, the levels of knockdown consistently achieved were in the range of the levels found elevated in tumor samples, that is, 3- to 5-fold. Both shRNAs sensitized U87MG cells, confirming the results of the screen (Fig. 2A). We observed that the depletion of SAT1 also sensitized D317MG and D54MG cell lines to IR, with dose-modifying factors at 10% survival of 1.14, 1.32, and 1.36 for U87MG, D317MG, and D54MG, respectively (Fig. 2B and C). To determine whether transient knockdown was also sufficient to sensitize cells to radiation, we performed siRNA in the D54MG line and found similar sensitization (Fig. 2D).

Tumors are comprised of cells with differing potentials to initiate or repopulate. The importance of such “tumor-initiating cells” in cancer therapy has become of peak interest due to their tendencies to be particularly resistant (5). We next evaluated the effect of SAT1 knockdown on radiation response of primary GBM neurosphere cell lines (GBM0821 and GBM0913) that are enriched in tumor-initiating cells (19) in a 3D neurosphere colony assay. We again observed sensitization to radiation following SAT1 depletion; GBM0821 by a factor of 1.70, and GBM0913 by a factor of 1.35 (Fig. 2E and F). Together, the results show that inhibition of SAT1 is sufficient to sensitize multiple tumor cell lines to radiation.

Clonogenic capacity in vitro is often used as a surrogate for tumorigenic potential in vivo. Besides observing increased radiosensitivity in SAT1-depleted neurospheres, we also found decreased in clonogenic capacity absent of radiation. To determine whether SAT1 affects tumor growth, nude mice were injected intracranially with luciferase-expressing SAT1 knockdown GBM0821 and GBM0913 neurospheres and compared with shGFP knockdown cells. Tumor growth was monitored using bioluminescent imaging. Over a 80- to 90-day period, we found that mice injected with control cells developed tumors more rapidly than mice injected with SAT1 knockdown cells from both neurosphere lines (Fig. 3A and B). Kaplan–Meier curves demonstrated significantly increased survival of animals with shSAT1 tumors compared with controls (Fig. 3C and D). In the GBM0913 line, control animals survived an average of 65.5 days after injection, whereas the shSAT1 animals survived 83.0 days (P = 0.0047). In the GBM0821 line, control animals survived an average of 47.0 days, and the shSAT1 animals survived 77.0 days (P = 0.0064). To determine whether eventual tumor growth in the shSAT1 neurosphere lines correlated with regained expression of SAT1, tumors were excised and subjected to qRT-PCR for SAT1. We found, however, that SAT1 knockdown was maintained in tumor samples over the course of the experiment, suggesting that SAT1 deficiency limited the growth of the tumor, rather than inducing a selective pressure against SAT1 knockdown (Supplementary Fig. S2A and S2B). Thus, elevated SAT1 expression in GBM promotes tumorigenesis in addition to reducing radiosensitivity.

To test whether SAT1 depletion can sensitize tumors in vivo, transient knockdown in U87MG tumors was achieved by intratumoral injection of siRNA using cationic lipid-based nanoparticles (14). Subcutaneous U87MG tumors 100 to 200 mm³ were injected with 500 pmol/L of siSAT1 or RISC control packaged in 1-aminoethyl)iminobis[N-(oleicylcysteinyl-1-amino-ethyl)-propionamide] (ECO), and subsequently irradiated 48 hours later with 8 Gy. Tumors were measured twice weekly, until the tumors reached ≥1.5 cm³. As can be seen in Fig. 3E, whereas transient knockdown of SAT1 had no effect on tumor growth, and 8 Gy IR had a modest effect on tumor growth, the combination of siSAT and radiation led to the most significant delay on tumor growth. Likewise, Kaplan–Meier survival curves of the animals demonstrated a significant benefit to siSAT1 and radiation over radiation alone (Fig. 3F). Together, the data argue that the sensitization of GBM cells in vitro exists in vivo as well.

Polyamine-induced DNA compaction/aggregation is known to be a means of radioprotection (8, 10, 20). Thus, how SAT1, a rate-limiting enzyme of polyamine catabolism, would protect GBM cells from radiation is nonintuitive. To gain insight into the mechanism of SAT1-mediated radioprotection, we determined by mass spectrometry the levels of polyamines in control and SAT1 knockdown cells. Surprisingly, stable knockdown of SAT1 did not result in increases in steady-state levels of spermine or spermidine (Fig. 4A). One explanation for this observation might be that stable knockdown cells acclimate to decreased SAT1 gene expression, as enzymes in the polyamine pathway are known to be tightly regulated by feedback mechanisms (21). To determine whether decreased catabolic enzyme expression (i.e., SAT1) was accompanied by decreased anabolic enzyme levels, we measured the expression of the rate-limiting regulator of polyamine synthesis, ornithine decarboxylase 1 (ODC1; ref. 21). We found that ODC1 expression mirrored SAT1, displaying significant repression with both SAT1 shRNAs (Fig. 4B). Thus, global alterations in polyamine levels are not evident in stable SAT1 knockdown cells, and therefore do not correlate with radiosensization following SAT1 depletion.

We next tested whether alterations in the cell cycle could explain increased sensitivity to radiation by increasing the fraction of cells in radiosensitive phases (e.g., G₁). We assessed the cell-cycle distribution of D54MG cells following stable knockdown of SAT1, but found no significant differences (Fig. 5A). In addition, SAT1 expression levels were not cell-cycle dependent (Fig. 5B). In contrast, when we assessed the sensitivity of cells sorted by flow cytometry using Hoechst 33342 (Supplementary Fig. S3), we found that S phase and G₂–M phase cells were uniquely sensitized to radiation following knockdown, whereas G₁ phase cells were unaffected (Fig. 5C). S phase cells displayed a dose-modifying factor of 1.90, at 10% survival; G₂–M cells displayed a factor of 1.86.

HR is the predominant DSB repair pathway in G₂–M and S phases of the cycle due to the presence of homologous sister chromatids after replication (22). Our findings of radiosensitization in G₂–M and S phases led us to investigate the effect of SAT1 on HR using an established MCF7-based HR reporter assay (DR-GFP; ref. 23). SAT1-deficient cells demonstrated a 54% decrease in HR compared with shGFP cells (P = 0.012; Fig. 5D). To test whether SAT1 knockdown would sensitize cells to an S phase–specific DNA-damaging agent, we treated D54MG cells with physiologically relevant concentrations of the topoisomerase I inhibitor Camptothecin (CPT). Chronic 10 nmol/L CPT exposure reduced colony survival in control cells by 10-fold; in contrast, CPT exposure led to a 1,000-fold reduction in SAT1 knockdown colony survival (Fig. 5E). Thus, SAT1 depletion sensitizes cells to IR and S phase agents by inhibiting HR.

To gain insight into the mechanism of regulation of HR by SAT1, we measured DNA damage after IR by comet assay in shGFP and shSAT1 D54MG cells. We found basal levels of damage and induced levels of damage after 10 Gy were similar. After 6 hours, however, SAT1-depleted cells demonstrated a marked decrease in the ability to repair compared with control (Fig. 6A). The HR protein BRCA1 was then used to identify HR foci in U87MG and D54MG SAT1 knockdown and control cells at 0 or 6 hours following exposure to IR (Fig. 6B). Although control cells displayed a potent induction of BRCA1 foci, SAT1 knockdown cells from both the U87MG and D54MG lines exhibited diminished capacity to produce foci. Both the number of cells with foci (64% vs. 16% for U87MG shGFP vs. shSAT1; and 66% vs. 32% for D54 shGFP vs. shSAT1), and the number of foci per cell were reduced (Fig. 6C). As the proportion of cells in S and G₂–M phases were unchanged between control and SAT1 knockdowns, the data suggest that SAT1-deficient cells have reduced HR signaling.

We next asked whether HR pathway components are altered in SAT1 knockdown cells. Western blotting revealed that BRCA1 itself is severely diminished in SAT1-depleted cells (Fig. 6D). In contrast, expression levels of MRE11, NBS1, and Rad51 remained unchanged. As BRCA1 could be regulated at various points, we examined mRNA and found consistent reductions of expression of BRCA1 after SAT1 depletion, suggesting regulation of BRCA1 by SAT1 is at the transcriptional level (Fig. 6E).

Increased H3 acetylation at the BRCA1 promoter has been reported to regulate BRCA1 transcription (24). Recent studies have demonstrated a link between elevated levels of spermidine and histone deacetylation in aging yeast (25). We hypothesized that alterations in SAT1 levels may affect BRCA1 transcription through regulation of chromatin access via histone acetylation. We first assessed global histone H3 acetylation on cell lysates and found in both U87MG and D54MG cell lines that SAT1 knockdown resulted in decreased N-terminal H3 acetylation (Fig. 6G). We next assessed the levels of acetyl-H3 at the BRCA1 promoter via ChIP and found similarly that the knockdown populations exhibited decreased levels of acetylated H3 localized to the BRCA1 promoter. In contrast, the β-actin promoter displayed mild increases in H3 acetylation (Fig. 6F). To determine whether reversing the reduction in histone acetylation would restore BRCA1 levels, we treated D54MG shGFP and shSAT1 cells with the HDAC (histone deacetylase) inhibitor TSA and found increased H3 acetylation and induction of BRCA1 mRNA (Fig. 6H). Thus, the data suggest that SAT1 promotes acetylation of histone H3 on the BRCA1 promoter, thereby inducing BRCA1 expression and facilitating HR repair.

To validate the concept that SAT1 mediates poor outcome and potentially response to therapy through BRCA1 expression, we sought to determine whether SAT1 and BRCA1 expression correlate in the Rembrandt data presented in Fig. 1D. Indeed, SAT1 and BRCA1 expression correlate with statistical significance (P = 0.0118; Fig. 7A), and BRCA1 expression revealed an outcome benefit for patients with lower expression as opposed to higher expression (P = 0.0001; Fig. 7B). Together, the clinical data support the importance of both SAT1 and BRCA1 in predicting patient outcomes, and suggest BRCA1 is at least partially responsible for the radioresistance of SAT1-overexpressing tumors.

In the present study, we have identified a novel mechanism of radioprotection in glioblastoma linked to the frequent overexpression of the polyamine catabolic enzyme SAT1. SAT1 is overexpressed almost uniformly in GBM tumor samples compared with normal brain, as assessed bioinformatically in the Oncomine and TCGA databases. Depletion of SAT1 radiosensitized multiple adherent cell lines and neurosphere lines, and tumors in mice. Mechanistically, we found that SAT1 promotes HR by controlling BRCA1 expression and BRCA1 foci following damage to DNA. We elucidated a novel function for SAT1 as a mediator of histone H3 acetylation controlling BRCA1 gene expression. Together, our findings highlight a previously unappreciated role for the polyamine metabolic pathway in regulating radiation responses in tumors, and identify a potential target for radiosensitizers to improve therapeutic efficacy.

Polyamines, including putrescine, spermine, and spermidine, are ubiquitous cellular constituents present in millimolar quantities. As positively charged small molecules, polyamines bind negatively charged macromolecules throughout cells, including nucleic acids and proteins, and regulate many processes (26). Polyamines promote protection to IR by inducing DNA compaction, producing a less susceptible target to direct radiation damage (8, 9, 27). SAT1, however, is the rate-limiting enzyme in polyamine catabolism, driving acetylation, and subsequent degradation or excretion of polyamines. Indeed, acetylation has been shown to result in removal of polyamines from chromatin, leading to chromatin relaxation and opening (28). Thus, existing data do not provide a mechanistic explanation for the role of SAT1 in radioprotection.

A recent proteomics study drew a connection between polyamine acetylation in radioresponse, finding acetyl-spermidine as a blood-borne biomarker in mice exposed to whole-body radiation (29). Aged mice showed a diminished ability to induce acetyl-spermidine following exposure, suggesting that aging-related decreases in DNA repair correlate with reduced polyamine catabolism. Although the mechanism behind the observations was not fully uncovered, the data intriguingly suggest that acetylated polyamines may provide a noninvasive marker of radiation responses.

To assess the effect of SAT1 expression on homeostatic polyamine levels in our system, we quantified cellular polyamines and found negligible differences. Although not surprising that cells acclimate to decreased SAT1 by reducing expression of ODC1 via described feedback loops (21), the data suggest that regulating global polyamine levels is not the mechanism of radioprotection by SAT1. Recent data have demonstrated, however, that controlling localized polyamine content is also an important function of SAT1. SAT1 has been found to modulate cell migration through an interaction with α9β1 integrin, localizing polyamine catabolism to membrane-bound potassium channels. Because spermine and spermidine are potassium channel blockers, the association of SAT1 with α9β1 in focal adhesions at the leading edges of migrating cells allows ion channel regulation (30). Thus, our data do not rule out that controlling localized polyamines at damage sites may be a mode by which SAT1 can affect radiation response.

Alternatively, the findings point to a novel mechanism of SAT1 in regulating histone acetylation. Excess polyamines have been recently shown to prolong lifespan via elevated autophagy due to alterations in histone acetylation and subsequent induction of autophagy genes (25). Furthermore, depletion of polyamines led to hyperacetylation of histone H3, generation of reactive oxygen species (ROS), and ultimately necrosis. Conversely, spermidine treatment caused deacetylation of histone H3 through inhibition of histone acetyltransferases. Conceptually, the findings agree with ours, that by decreasing polyamines, SAT1 would promote acetylation of H3 and subsequently alter gene expression. In our studies, however, the importance of polyamines was unclear. As an acetyl-transferase, it is unknown whether SAT1 can also target proteins, as no specific targets have been identified beyond an autoacetylation activity (7). In yeast, the hpa1 and hpa2 genes, though structurally unrelated to SAT1, have been found to be capable of acetylating both polyamines and histones (31). In vitro, we were unable to detect a direct acetylation activity of purified SAT1 on purified histone H3, however, leaving open the question of a direct or indirect effect of SAT1 on histones. Alternatively, the observation that the HDAC inhibitor reversed the acetylation status of H3 and rescued BRCA1 levels may suggest that SAT1 can function as an HDAC inhibitor. Future studies will be necessary to bear out this possibility.

Polyamine analogues have long been a focus of development of anticancer agents (32, 33). The mechanism of cell death by analogues is due to potent induction of SAT1 (>1,000-fold), leading to polyamine depletion and excess production of ROS (21). Ectopic overexpression of SAT1 is also toxic due to inhibition of general protein translation (34). To avoid complications of nonphysiologic alterations in SAT1 levels, we focused on shRNAs that reduced SAT1 levels to the range observed in normal brain (3- to 5-fold reductions in the GBM lines). In this context, our data suggest that decreasing SAT1 may have therapeutic benefit. Furthermore, sensitizing brain tumors with an SAT1 inhibitor may have relevance in combination with Parp inhibitors, which take advantage of alterations in the BRCA1 pathway and have recently been shown to sensitize brain tumor-initiating cells to radiation (35).

In summary, we have defined a novel function for the polyamine catabolic enzyme SAT1 in mediating HR repair in brain tumors through the epigenetic regulation of BRCA1. Our findings describe a new mechanism for SAT1 to regulate chromatin, and suggest that inhibition of SAT1 may sensitize brain tumors to radiation and increase therapeutic responses. Together, the findings contribute to our understanding of radioresistance in GBM, and may have yet even broader implications due to the overexpression of polyamine metabolic enzymes in a variety of other tumors (36).

No potential conflicts of interest were disclosed.

Conception and design: J. Zhang, S.M. Welford

Development of methodology: Y. Seo, J. Zhang, R Spina, E.E. Bar, M. Gujrati, Z.-R. Lu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Brett-Morris, B.M. Wright, Y. Seo, V. Pasupuleti, J. Lu, E.E. Bar, M. Gujrati, R. Schur

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Brett-Morris, B.M. Wright, S.M. Welford

Writing, review, and/or revision of the manuscript: A. Brett-Morris, Y. Seo, V. Pasupuleti, J. Lu, S.M. Welford

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B.M. Wright, R Spina

Study supervision: A. Brett-Morris, E.E. Bar, S.M. Welford

The authors thank Dr. Ravi Patel for colony-counting software. Core facilities of the Case Comprehensive Cancer Center, supported by P30CA43703, were used: Radiation Resources, Cytometry and Imaging Microscopy, Small Animal Imaging.

This work was supported by grant 119999-IRG-91-022-18-IRG from the American Cancer Society and by grant P30CA43703 from the Case Comprehensive Cancer Center. J. Zhang is supported by NCI R01CA154625.

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.