Changes in Pyruvate Metabolism Detected by Magnetic Resonance Imaging Are Linked to DNA Damage and Serve as a Sensor of Temozolomide Response in Glioblastoma Cells

Temozolomide (TMZ) is a chemotherapeutic DNA–methylating agent useful in the treatment of glioblastoma (GBM; refs. 1, 2). DNA O⁶-methylguanine (O6MG) lesions created by drug exposure mispair with thymine in the first cycle of DNA replication, leading to activation of the DNA mismatch repair system (3–5). The removal of the thymine opposite O6MG and subsequent reinsertion following repair resynthesis leads to cycles of futile mismatch repair. This in turn ultimately leads to DNA single-strand breaks (SSB), activation of sensors of DNA damage, the generation of DNA double-strand breaks (DSB), and delayed cell death by a combination of senescence and mitotic catastrophe (6–10). Because of its limited side effects and proven efficacy, TMZ is a component of the standard therapy for a variety of human brain tumors, and is frequently combined with other experimental agents (2).

Despite its ability to prolong disease-free survival, the response to TMZ-based therapies has proven difficult to monitor in patients. At present, standard imaging techniques are used to monitor the size of contrast-enhancing tumor following initiation of therapy, although for reasons related to difficulties in distinguishing tumor from necrosis and scarring, this approach often proves suboptimal (11, 12). Given that many patients, especially those with low-grade gliomas, can remain on TMZ for long periods of time (13), a more effective way to monitor drug efficacy and the early stages of drug failure in real-time would allow for better patient stratification as well as for the initiation of alternate approaches in cases of drug failure.

It has been known for more than 50 years that tumors metabolically differ from normal cells (14), an observation that has recently been applied to the noninvasive study of tumors by carbon-13 (¹³C) metabolic imaging. Tumors frequently have high levels of lactate dehydrogenase A (LDH-A; which preferentially converts pyruvate to lactate) and NADH, the cofactor required for LDH-A activity, both of which may explain the high levels of lactate found in many tumors (15, 16). Changes in the expression of several other tumor-specific metabolic enzymes have also been noted, however, including a shift in expression from pyruvate kinase (PK) M1 (which complexes with multiple glycolytic enzymes and facilitates the shunting of pyruvate into the citric acid cycle) to the less active M2 form of the enzyme, which also may contribute to increased pyruvate to lactate conversion (17–19). On the basis of these observations, the administration of hyperpolarized compounds, most typically pyruvate, to tumor-bearing subjects, followed by the real-time monitoring of conversion of the substrate to various labeled metabolites using ¹³C 3D magnetic resonance spectroscopic imaging (MRSI) should allow for the localization of malignant tissue (20). Studies along these lines have proven promising, and further work suggested that the approach could also be used to study drug response because cells undergoing growth arrest or apoptosis convert less hyperpolarized pyruvate to lactate (21–24). More recently, the increased conversion of hyperpolarized pyruvate to lactate was shown to precede the onset of tumor growth, and also was reversed before tumor regression, at least in a myc-driven murine model of liver cancer (25). These studies suggest that hyperpolarized ¹³C MRSI may be a promising way to monitor tumor growth and perhaps tumor drug response.

We previously used hyperpolarized ¹³C MRSI monitoring to show that human GBM xenografts exhibit an increase in hyperpolarized pyruvate to lactate conversion relative to normal brain (26, 27), and that TMZ exposure decreased this conversion in a time frame that preceded histologic and radiographic response (28). Although these findings suggested that pyruvate metabolism might be linked to TMZ response and could serve as a biomarker of TMZ action, it was not clear whether the metabolic response was unique to drug-sensitive cells, nor was the basis for the metabolic change apparent. To better address these issues, we generated two pairs of isogenic GBM cells that differ only in expression of O6MG DNA methyltransferase (MGMT), the protein that removes TMZ-induced O6MG lesions before they can mispair and lead to DNA damage, and as such serves as the primary determinant of TMZ sensitivity (29–31). Using these cells, we show that the TMZ-induced suppression of the conversion of hyperpolarized pyruvate to lactate is uniquely linked to the TMZ-induced DNA-damage response via decreased PKM2 expression and intracellular lactate levels, and as such serves as an early monitor of TMZ response in brain tumor cells.

U87MG human GBM cells, and the same cells expressing MGMT P140K (U87 + MGMT), or shRNAs targeting PKM2 (U87 + shRNA PKM2) were obtained and cultured as described previously (31, 32). Suppression of Chk1 protein (>80%) by transfection of siRNA (Santa Cruz Biotechnology) was verified by Western blotting. G55 is a human GBM cell line passaged through nude mice and reestablished as a stable xenograft (33), and was provided by the UCSF Brain Tumor Research Center. The G55 cells form invasive intracranial tumors in rodents that are more characteristic of primary human GBM (33). The identity of both cell lines was verified by DNA fingerprinting (Promega Powerplex 16) before initiation of the studies.

Tumor cells were implanted intracerebrally into male athymic rats (median weight 280 g) as described previously (26–28). Animals received either oral vehicle (Ora-Plus; Paddock Laboratories), oral TMZ (100 mg/kg), intraperitoneal O⁶-benzylguanine (BG, 50 mg/kg), or BG followed 2 hours later by TMZ when the tumor as monitored by T₂-weighted axial images was at least the size of one spectroscopic voxel (approximately 86 mm³, typically 9–15 days after tumor cell implantation). All animals underwent ¹³C and ¹H imaging before treatment (D0, baseline), and 1 (D1) and/or 2 (D2) days following treatment.

The polarization procedure and in vivo imaging have been described previously (17, 24, 25, 27, 34–36). In brief, ¹³C imaging data were acquired using a compressed sensing or fully sampled ¹³C 3D MRSI sequence (24, 25) at 20 seconds from the start of the injection of approximately 2.5 mL hyperpolarized [1-¹³C]-pyruvate through the tail vein. The compressed sensing data were acquired over 18 seconds, and the fully sampled data over 17 seconds. The injection started approximately 10 seconds after dissolution and lasted 10 seconds. For tumor volume evaluation, T₁-weighted anatomical images were obtained in the axial plane after the injection of 0.1 mL Gadolinium (Gd)-DTPA (approximately 0.2 mmol/kg; ref. 27).

The methods for processing ¹³C MRSI data have been described previously (27, 36, 37). The 3D volume of the contrast-enhancing lesion was calculated from the axial T₁ post-Gd slices as described previously (36). At each time point a percentage of tumor volume change from baseline was calculated, with the last time point being a composite of values derived from multiple paired animals at the termination of the experiment 3 to 7 days after TMZ exposure.

At D1 or D2 after drug treatment, animals were anesthetized (1%–2% isoflurane) and perfused with 10% neutral-buffered formalin (40 mL). The brains were then removed, fixed in phosphate-buffered 10% formalin, dehydrated by graded ethanols, and embedded in Paraplast Plus wax (McCormick Scientific). Tissue sections were incubated with rabbit monoclonal phospho-Chk1 (Ser345) or polyclonal cleaved caspase-3 (Asp175) antibodies (Cell Signaling Technology) with antigen retrieval. All immunohistochemistry assays were performed on the Benchmark XT (Ventana Medical Systems) using the UltraView detection system.

Approximately 3 × 10⁷ U87 or U87 + shRNA PKM2 cells were grown on Biosilon microcarrier beads (Nunc) for 48 hours, then loaded and analyzed in a 10 mm NMR tube connected to an NMR-compatible perfusion system, as previously described (23, 38).

Whole-cell lysates (30 μg protein) were subjected to Western blot analysis using antibodies against Chk1, MGMT, PKM2, and β-actin (all Cell Signaling Technology; ref. 32) with detection by a horseradish peroxidase–conjugated secondary IgG (Santa Cruz Biotechnology) using enhanced chemiluminescence Western blotting detection reagents (Amersham Pharmacia Biotech).

LDH activity and NADH levels were assessed from the D2 resected brains of control and TMZ-treated U87 tumor-bearing rats. Brains were rapidly removed after euthanasia, immediately snap-frozen in liquid N₂, and stored at −80°C. For LDH activity, K_m and V_max values were determined by fitting initial rates of NADH consumption (absorbance at 340 nm) and pyruvate concentrations to a Lineweaver–Burke plot (39). Levels of NADH were determined in tumor extracts using a spectrophotometric enzymatic cycling assay kits (BioVision) according to the manufacturer's instructions (40). In brief, tumor tissue was homogenized, centrifuged (14,000 × g, 5 minutes) and heated (60°C, 30 minutes) to selectively decompose NAD⁺ in the supernatant. The samples were then diluted 1:10 to fall in the linear range for a standard curve, and mixed with NADH reaction buffer. After 15 minutes of incubation at room temperature, changes in absorbance at 450 nm caused by the reduction of the reporter molecule thiazolyl blue were measured on a M200 Tecan spectrophotometer (Tecan Group) and compared with a standard curve established using known NADH concentration.

Intracellular PK activity, as well as concentrations of pyruvate and lactate were measured using PK activity, pyruvate, or lactate assay kits, respectively (BioVision).

To first determine whether TMZ-induced changes in pyruvate metabolism were related to drug response, two pairs of isogenic cells differing only in MGMT expression were used. TMZ-sensitive, MGMT-deficient GBM cells (U87 cells and G55 cells preincubated with the selective MGMT-depleting agent BG; ref. 41) exhibit increased levels of DNA SSB within 1 day of an IC₅₀ TMZ exposure, accompanied by increased levels of the DNA-damage response protein pChk1, and the appearance of DNA DSB beginning 2 days after TMZ exposure (31). Isogenic TMZ-resistant, MGMT-proficient GBM cells (U87 cells expressing a construct encoding MGMT, and G55 cells) in contrast, exhibit no increase in DNA SSB, pChk1, or DSB at 1 or 2 days following TMZ exposure (31).

In the first set of studies, control and MGMT-modified U87 GBM cells were intracranially implanted into rats and allowed to form tumors. Animals were then treated with vehicle or TMZ (100 mg/kg) and 24 or 48 hours later were administered [1-¹³C]-pyruvate and monitored for conversion of the substrate to lactate by ¹³C MRSI, for markers of the DNA-damage response, and for tumor growth. Nondrug-treated cells (U87 vehicle and U87 + MGMT vehicle) as well as the TMZ-resistant drug-treated U87 + MGMT + TMZ cells (31) exhibited an increase in pyruvate to lactate conversion both 1 and 2 days after TMZ treatment (Fig. 1A). The TMZ-sensitive drug-treated U87 cells (U87 + TMZ), however, exhibited decreased conversion of pyruvate to lactate (compared with pretreatment D0 tumors) beginning 1 day after TMZ exposure (Fig. 1A). The effect of TMZ on pyruvate to lactate conversion in these cells was consistent temporally with the appearance of cells that stained positively for activated pChk1 (Fig. 1B), but preceded the effect of the drug on tumor growth, which was not apparent 1 to 2 days after TMZ exposure in any cell group, but was apparent in the U87 + TMZ cells at the termination of the experiment 3 to 7 days after TMZ exposure (Fig. 1C; ref. 28). These results suggested that TMZ-induced changes in pyruvate to lactate conversion were related to the formation and/or response to TMZ-induced O6MG, and TMZ response.

To verify and expand on these results, MGMT-proficient, TMZ-resistant G55 GBM cells were intracranially implanted into rats, allowed to form tumors, treated with vehicle, the selective MGMT-depleting agent BG, TMZ only, or BG + TMZ, and similarly examined for effects on drug action and metabolic response. As with the tumors generated from the intracranially implanted U87 series of cells, nondrug-treated G55 cells (G55 + vehicle) as well as the G55 cells incubated with BG (G55 + BG) or TMZ alone (G55 + TMZ; ref. 31) exhibited an increase in pyruvate to lactate conversion both 1 and 2 days after TMZ treatment (Fig. 2A). The BG-treated, MGMT-depleted G55 cells (G55 + BG + TMZ), however, when exposed to TMZ, exhibited a decreased conversion of pyruvate to lactate (compared with pretreatment D0 tumors) beginning 1 day after TMZ exposure (Fig. 2A). The effect of TMZ on pyruvate to lactate conversion in these cells appeared coincident with the appearance of pChk1 staining at 1 day after TMZ exposure in the TMZ-sensitive G55 + BG tumors (Fig. 2B), but preceded the effect of the drug on tumor growth, which, consistent with the mechanism of action of the drug, was not apparent 1 to 2 days after TMZ exposure in any group (Fig. 2C), but was apparent in the G55 + BG + TMZ tumors, which at the termination of the experiment 3 to 7 days after TMZ exposure were 43% smaller than at 2 days, as opposed to G55 + TMZ tumors, which nearly doubled in size over the same time period. These results show that the effect of TMZ on pyruvate metabolism is directly related to the absence of MGMT, the generation of O6MG-related DNA damage, and/or the cellular response to these lesions, all of which occur before the onset of TMZ-induced growth arrest and cell death.

Reductions in the conversion of ¹³C pyruvate to ¹³C lactate in in vivo imaging studies of drug-treated tumors have been ascribed to a variety of factors, including loss of cellularity, apoptosis-induced PARP activation and depletion of NADH levels, decreased LDH-A expression, and loss of LDH activity. In the present study, the decrease in conversion of pyruvate to lactate in response to TMZ occurred within 1 day of TMZ exposure in vivo, long before any changes were noted in tumor growth and long before TMZ-induced O6MG lesions are converted to DNA mismatches and DNA DSB in these cells (31). Similarly, the effects of TMZ on pyruvate to lactate conversion were noted in drug-sensitive cells in the absence of any effect on apoptosis because control and TMZ-treated U87 tumors exhibited equally low levels of caspase-3 staining 2 days after TMZ exposure (Fig. 3A), consistent with the reported inability of these cells to undergo apoptosis in response to TMZ (6). Similarly, although TMZ-sensitive GBM cells could be expected to activate PARP following TMZ exposure in response to TMZ-induced DNA strand breaks (42), no changes in the levels of NADH (Fig. 3B) or LDH activity (Fig. 3C) were noted in U87 tumors 2 days after TMZ exposure as both the LDH K_m and V_max were not significantly different between control and TMZ-treated U87 tumor cells at the time (2 days after TMZ) at which significant differences in pyruvate to lactate conversion were noted (Fig. 1A). These results show that the TMZ-induced changes in pyruvate to lactate that are unique to TMZ-sensitive cells occur before the onset of growth arrest or cell death, and whereas related to TMZ-induced DNA damage and the damage response, appear unrelated to NADH levels and changes in LDH activity.

PKM2 is overexpressed in GBM (32) and in its tetrameric active conformation physically associates with other glycolytic enzymes (hexokinase, glucose 6-phosphate dehydrogenases, glyceraldehyde 3-phosphate dehydrogenase) that are part of the complex that controls the conversion of glucose to pyruvate and the ultimate fate of pyruvate. PKM2 has also been shown to be downregulated in cells exhibiting alterations in pyruvate to lactate conversion (25), suggesting that its regulation may contribute to the TMZ-induced changes in pyruvate metabolism noted in this study. To address this possibility, we monitored the effects of TMZ on PKM2 expression, PK activity, and pyruvate metabolism in TMZ-sensitive and TMZ-resistant U87 and G55 cells. Exposure of MGMT-deficient U87 or G55 + BG cells, but not MGMT-proficient U87 + MGMT or G55 cells, to TMZ decreased PKM2 expression (Fig. 4A) as well as PK activity (Fig. 4B) in a time frame (1 day following TMZ exposure) consistent with the decrease in pyruvate to lactate conversion in the same cells in vivo (Figs. 1A and 2A). These results suggest that the DNA-directed effects of TMZ on PKM expression may drive alterations in metabolism. To address this possibility more fully, we used two different shRNAs to suppress PKM2 expression in U87 cells and then monitored the effects on pyruvate metabolism. As shown in Fig. 4C, shRNA-mediated suppression of PKM2 expression resulted in a decrease in PK activity similar to that noted following TMZ exposure, as well as decreases in steady-state levels of both pyruvate and lactate in U87 and G55 cells (Supplementary Fig. S3; ref. 32). More importantly, the TMZ-sensitive U87 cells in which PKM2 levels and PK activity were suppressed also exhibited the same significant decrease in the in vitro conversion of hyperpolarized [1-¹³C]-pyruvate to lactate (63.5% ± 4.4% of U87 control, Student t test P < 0.01, n = 3; Fig. 4D and E), as was noted in TMZ-treated U87 cells. Similarly, in vitro levels of intracellular pyruvate and lactate were also significantly reduced in TMZ-treated U87 cells, and TMZ + BG-treated, MGMT-depleted G55 cells (Supplementary Fig. S4). These results suggest that the DNA-directed actions of TMZ lead to an early suppression of PK expression, PK activity, and intracellular lactate levels. The decreased intracellular lactate levels in turn are associated with, and may be responsible for (21), the decreased conversion of hyperpolarized pyruvate to lactate noted in TMZ-treated tumors.

Because TMZ-induced changes in PKM2 expression and pyruvate metabolism were noted within 1 day of TMZ exposure in MGMT-deficient cells, we considered the possibility that the DNA-damage response was linked to these metabolism-related changes. Furthermore, because the onset of Chk1 activation in drug-treated, TMZ-sensitive cells noted here and in previous studies (31) temporally matched that of the changes in metabolism, we examined the specific role of Chk1 in controlling this phenomenon. To do so, U87 and G55 cells were infected with siRNA constructs targeting Chk1, after which effects on PKM2 expression and PK activity were monitored. As shown in Fig. 5A, introduction of either of two different Chk1 siRNAs reduced levels of Chk1 by approximately 80% relative to scramble control cells in both cell lines. Introduction of scramble siRNA into U87 or G55 + BG cells did not significantly change PKM2 protein levels, whereas exposure of these cells to an IC₅₀ concentration of TMZ significantly decreased levels of PKM2 (Fig. 5B and consistent with data in Fig. 4A). This degree of TMZ-induced PKM2 suppression was significantly reduced by Chk1 suppression (last column vs. adjacent two columns, Fig. 5B). Consistent with these observations, exposure of control or scramble siRNA U87 cells, or of G55 + BG or G55 + BG + scramble siRNA cells to TMZ significantly decreased PK activity, an effect that was also lessened by Chk1 suppression (Fig. 5C). These studies show that the Chk1 activation caused by TMZ-induced O6MG damage also contributes to the decreases in PKM2 expression, PK activity, and intracellular lactate levels linked to changes in pyruvate metabolism.

Having shown that exposure of TMZ-sensitive, but not TMZ-resistant cells led to decreases in PKM2 expression and PK activity, and that these changes were sufficient to alter pyruvate metabolism, we also considered the possibility that these events were critical downstream components of the cytotoxic action of TMZ in addition to potential biomarkers of TMZ action. To address this possibility, we measured the cytotoxic potential of TMZ in drug-sensitive U87 cells and in the same cells in which levels of PKM2 were either genetically suppressed or in which PK activity was increased by virtue of overexpression of the highly active form of PK (PKM1). As shown in Table 1, whereas the suppression of PKM2 levels in TMZ-sensitive U87 cells that were shown to alter pyruvate metabolism modestly decreased the colony formation ability of the non TMZ-treated cells, it had no effect on TMZ-induced cytotoxicity. Similarly, although overexpression of PKM1 nearly doubled PK activity in U87 cells (32), it also had no significant effect on TMZ-induced suppression of clonagenicity. These results therefore suggest that, although TMZ-induced changes in pyruvate metabolism are linked to changes in PKM2 expression, PK activity, and ultimately TMZ-induced DNA adducts, these changes, although separable from the cytotoxic actions of the drug and not per se a contributor to the effects of TMZ on tumor cell clonagenicity, are a marker of TMZ activity (Fig. 5D).

¹³C MRSI is capable of monitoring the growth and regression of oncogene-driven tumors (22), and can provide biomarkers of the response of solid tumors to DNA-damaging agents and radiation (21). Here, we show that ¹³C MRSI measurements of hyperpolarized pyruvate to lactate conversion, via linkage to regulation of PKM expression and DNA damage, may be useful in measuring the early response of glioma to the commonly used chemotherapeutic-methylating agent TMZ.

A variety of studies have used ¹³C MRSI to examine the linkage between tumor growth, regression, drug response, and pyruvate metabolism, with varying results. Rapid changes in pyruvate metabolism were related to consumption of NADH and subsequent inhibition of LDH-A associated with apoptosis or necrotic cell death (21). Additional studies in doxorubicin-treated breast adenocarcinomas, however, showed that drug-induced suppression of pyruvate to lactate conversion could be separated from the cell death process, although the metabolic changes were still linked to consumption of NADH and subsequent inhibition of LDH-A associated with drug DNA damage and PARP activation (43). In the present study, the metabolic effects linked to TMZ sensitivity could be clearly separated from the cell death process and in turn linked to DNA damage, although the exact form of damage that triggers the change in metabolism noted can only be partially determined. The requirement for O6MG lesions, the timing of events, and the data from siRNA suppression studies, however, suggest that Chk1 activation may help link persistent DNA damage to the metabolic changes noted, thus making Chk1 important in both reacting to DNA damage and in providing a noninvasively detectable signal of cellular response.

If O6MG-induced DNA damage and signaling lead to changes in cellular metabolism, how is this accomplished? The present studies suggest that PKM2, a key regulator of cell metabolism known to be overexpressed in GBM may be involved. TMZ-induced DNA damage and Chk1 activation led to decreases in PKM2 levels and activity. Because PKM2 is regulated by a variety of events, including phosphorylation, PKM2 may be a target of Chk1 and as such directly linked to the DNA-damage response (44–46). Alternatively, the relatively large number of Chk1 substrates may activate pathways that alter PKM2 localization as well as absolute levels in a less direct manner. A number of other kinases, including Chk2, are also activated by TMZ in the time frame consistent with the TMZ-induced metabolic changes, and this degree of redundancy may explain the only partial reaquisition of PKM2 levels and activity following Chk1 knockdown in TMZ-treated cells. The TMZ-induced decrease in PKM2 levels was in turn associated with decreased levels of pyruvate and lactate at least where measured in vitro. Given the previously described connection between reduced intracellular lactate levels and decreased conversion of ¹³Cpyruvate to ¹³C lactate (21), it seems reasonable that the TMZ-induced decreases in intracellular lactate levels noted here resulted in the decreased pyruvate to lactate conversion in TMZ-treated animals. Although direct proof of this linkage would require more complex methods such as diffusion-weighted spectroscopy to measure intracellular lactate levels in vivo (47, 48), the present data clearly show that PKM2 plays a key role in linking the DNA-damage response to monitorable changes in pyruvate metabolism that in turn relate to drug efficacy.

In addition to identifying the link between TMZ-induced DNA damage and pyruvate metabolism, and defining the mechanism by which this linkage occurs, the present study also has clinical implications. Current monitoring of TMZ action relies on inaccurate and uncertain measures of tumor size gathered over long periods of time following drug exposure. The studies presented show that at least in a preclinical model of GBM, TMZ-induced changes in pyruvate metabolism can provide a much earlier and perhaps more reliable marker of tumor response. Detection of the TMZ-induced changes in metabolism do not rely on measurements of tumor size or contrast enhancement, and in fact appear to precede any drug-induced changes in these parameters. As such this approach may alleviate concerns about inaccuracies of tumor size measurements. In addition, TMZ-induced changes in metabolism are related to the same event (DNA damage) that leads to cell death, but do not appear to be part of the cell death process. As a result, direct drug-sensitivity information, rather than information indirectly inferred from predictors of DNA damage (such as MGMT promoter methylation; ref. 49), can be obtained rapidly, repeatedly, and in the actual tumor being treated within a day of drug exposure rather than days to weeks later. Clinically, this could allow more rapid decision making and could also limit the exposure of drug-insensitive tumors to mutagens such as TMZ. The latter point could be especially important in low-grade astrocytoma, which are frequently treated with TMZ for extended periods of time, and that can be driven to progression by TMZ-induced genetic alterations (50). Therefore, although metabolic changes may be driven by events directly linked to cell growth inhibition and cell death, the unique delayed action of TMZ may provide links between the DNA-damage response and cell metabolism that can be exploited for improved therapy.

No potential conflicts of interest were disclosed.

Conception and design: I. Park, J. Mukherjee, M. Ito, S.J. Nelson, R.O. Pieper

Development of methodology: I. Park, J. Mukherjee, K. Gaensler

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I. Park, J. Mukherjee, M.M. Chaumeil, L.E. Jalbert

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I. Park, J. Mukherjee, M.M. Chaumeil, L.E. Jalbert, S M. Ronen, S.J. Nelson, R.O. Pieper

Writing, review, and/or revision of the manuscript: I. Park, J. Mukherjee, M.M. Chaumeil, L.E. Jalbert, S.M. Ronen, S.J. Nelson, R.O. Pieper

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Gaensler

Study supervision: S.M. Ronen, R.O. Pieper

The authors thank Lynn DeLosSantos, Dr. Robert Bok, Kristen R. Scott, Pia Eriksson, and Loretta Chan for technical support, and Dr. Tal Raz for statistical analysis.

This study was supported by NIH grants CA171610, CA097257, and CA118816 (R.O. Pieper), CA130819 and EB004 (S.J. Nelson), and CA161545, CA172845, CA15491, and EB013598 (S.M. Ronen).

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.