|Home | About | Journals | Submit | Contact Us | Français|
In this review, the authors describe a novel mechanism for control of MYC expression that involves a four-stranded DNA structure, termed a G-quadruplex, amenable to small molecule targeting. The DNA element involved in this mechanism, the nuclease hypersensitive element III1 (NHE III1), is just upstream of the P1 promoter and is subjected to dynamic stress (negative superhelicity) resulting from transcription. This is sufficient to convert the duplex DNA to a G-quadruplex on the purine-rich strand and an i-motif of the pyrimidine-rich strand, which displaces the activating transcription factors to silence gene expression. Specific proteins have been identified, NM23-H2 and nucleolin, that resolve and fold the G-quadruplex to activate and silence MYC expression, respectively. Inhibition of the activity of NM23-H2 molecules that bind to the G-quadruplex silences gene expression, and redistribution of nucleolin from the nucleolus to the nucleoplasm is expected to inhibit MYC. The authors also describe the mechanism of action of Quarfloxin, a first-in-class G-quadruplex-interactive compound that involves the redistribution of nucleolin from the nucleolus to the nucleoplasm. G-quadruplexes have been best known as test-tube oddities for more than four decades. However, during the past decade, they have emerged as likely players in a number of important biological processes, including transcriptional control. Only time will tell if these odd DNA structures will assume the role of an established receptor class, but it is clear from the scientific literature that there is a dramatic increase in interest in this little-known area in the past few years.
G-quadruplexes and i-motifs are four-stranded DNA structures that can be derived from intramolecular or intermolecular folding of DNA single strands (Figure 1). Non-B-DNA duplex structures such as A-DNA and Z-DNA have had a long history, and in the case of Z-DNA, their biological significance has been controversial.1,2 In view of this, it is not surprising that the proposal of these four-stranded structures having functional significance in cells has been received with considerable skepticism, especially by biologists. A G-quadruplex structure was first identified in 1962 by Gellert and coworkers3 and then proposed by Sen and Gilbert in 1988 to have putative biological significance—in this case, in meiosis.4 Its existence in the telomeric regions of cellular DNA was firmly established more than a decade later using fluorescent antibodies in Stylonychia lemnae.5 It has been demonstrated that small molecule targeting of these structures in human telomeres causes inhibition of telomerase and disruption of the telomeric ends,6 suggesting that they might be therapeutic targets, as first proposed in 1989.7
Following the early interest in telomeric G-quadruplexes, it was shown that the purine-rich strand in the NHE III1 of the MYC promoter could form a G-quadruplex, and there was speculation that this structure could have functional significance.8 Fueling this was the observation that two single-stranded DNA-binding proteins, CNBP and hnRNP K, could bind to this same element.9–11 Thus, a single-stranded element could be a precursor to the formation of secondary DNA structures (G-quadruplexes and i-motifs on the G-rich and C-rich strands, respectively).12 Additional support for this idea came from cellular experiments in which transcription of MYC and hTERT was inhibited upon addition of the G-quadruplex-interactive compound TMPyP4, whereas the positional isomer TMPyP2 (Figure 2A), which weakly bound G-quadruplexes, had a much lesser effect.13 However, the most convincing evidence came from experiments in which luciferase reporter and cellular experiments with two Burkitt lymphoma cell lines were used to evaluate the transcriptional effect of targeting the purported G-quadruplex in the NHE III1.14 In these cases, suppression of luciferase or MYC expression by TMPyP4 depended on either the wild-type sequence, which was able to form a stable G-quadruplex, or the preservation of the G-quadruplex-forming element on chromosome 14 after translocation (Figure 2B; see later). This established the proof of principle that small molecule stabilization of the G-quadruplex in the MYC NHE III1 would silence gene expression and, furthermore, that the G-quadruplex was the silencer element. However, additional experiments were still required to convince a justifiably skeptical audience based on the still largely unfulfilled promise of Z-DNA after three decades of work.1
Additional questions that needed to be addressed relating to these unique structures in the MYC promoter were as follows:
We have directly answered questions 1 and 2, and we address this evidence, along with strategies and future challenges in modulating MYC gene expression, in the remainder of this review.
The transcriptional control of MYC gene expression via the far upstream element (FUSE) and NHE III1, using an interplay of transcriptional factors and dynamic consequence of transcriptionally induced negative superhelicity, is a first-in-class example of a new level of complexity in transcriptional control and is probably a common feature of many other eukaryotic genes.15,16 This concept was developed in David Levens’s laboratory at the National Cancer Institute.16,17 Because this pioneering work on the FUSE is described in another review in this issue (see Levens88), it will not be recapitulated here, except to point out that this is the “cruise control” component of the overall system, whereas the NHE III1 is the “on/off switch” for MYC. The overall features of this on/off switch centrally involve the formation and dissipation of the G-quadruplex/i-motif structures and are controlled by nucleolin and NM23-H2, respectively, with transcriptionally induced negative superhelicity (Figure 3A). In the following subsections, each feature is described as it relates to the need of the cell to activate or silence gene transcription.
Until recently, all the biophysical and chemical characterization of the G-quadruplex and i-motif in the MYC NHE III1 had been carried out in the context of a single-stranded DNA template ex vivo.18 The NHE III1 contains six runs of three or more contiguous guanines and cytosines (Figure 1). In principle, only four runs are required for either secondary structure, and it was initially assumed that the five guanine runs having one or two base loops at the 5′ end were involved, rather than an alternative arrangement including the sixth run at the 3′ end, which is separated from the other runs by a five-base-pair intervening loop sequence. DMS footprinting in the single-strand context revealed that runs 2–5 preferentially form a parallel-stranded G-quadruplex.19,20 The picture was less clear for the C-rich strand where, depending on conditions, ill-defined intramolecular or intermolecular species were obtained at acidic pH levels.21
Perhaps not too surprising, it was found that in the context of double-stranded DNA under conditions of negative superhelicity, the precise guanine and cytosine runs used in the formation of the G-quadruplex and i-motif of the MYC promoter18 were found to be different from those used from a single-stranded DNA template.19–21 The G-quadruplex in the MYC NHE III1 forms from guanine runs 1–4 only under conditions of negative superhelicity and requires the wild-type sequence (Figure 1). A single-base mutation in the center of guanine run 4 prevents G-quadruplex formation under these conditions.18 On the C-rich strand, five runs of cytosines at the 5′ end are used, with cytosine run 3 located entirely in one of the loops (see Figure 1). The major isomer formed, based on Br2 footprinting, is the 6:2:6 loop form.18 (Note that 6:2:6 refers to the number of bases that connect the two sets of cytosine+–cytosine base pairs that make up the i-motif. The bases in these loops are likely involved in stabilizing H-bonding and base-stacking interactions; thus, these loops are known as capping structures.) This result finally clarified the need for six, rather than four, runs of cytosines as a requirement for stable i-motif formation. Somewhat surprising, under conditions of negative superhelicity, the i-motif forms from duplex DNA under physiological conditions rather than the acidic pH conditions required in the single-stranded DNA context. This is probably due to the torque in the NHE III1 caused by the negative superhelicity that is relieved upon formation of the i-motif and G-quadruplex, with stabilizing capping structures in the 6:2:6 loops. Because the i-motif uses the five 5′-runs of cytosines, it is offset from the G-quadruplex (Figure 1). S1 nuclease experiments indicate that there are flanking regions of single-stranded character.18
Nucleolin is 100-kDa multifunctional nucleolar phosphoprotein that plays a number of important roles in cells, including ribosome biogenesis, transcriptional regulation, chromatin decondensation, cell proliferation, and differentiation.22–26 Because of its modular structure, it can adapt to interact with nontraditional forms of RNA and DNA,27,28 including G-quadruplexes.29–31 Recent studies have demonstrated that nucleolin selectively binds to parallel-stranded G-quadruplexes and, in particular, to the MYC G-quadruplex, where it facilitates folding from the single-stranded form and then stabilizes it upon binding.32 Furthermore, ChIP analysis has shown that nucleolin binds to the NHE III1, presumably as the G-quadruplex structure, and that nucleolin overexpression is able to inhibit Sp1-induced MYC transcriptional activation.32 The initial observation that nucleolin binds to the MYC G-quadruplex rather than single-stranded or duplex DNA, alongside the demonstration that it binds to the NHE III1 in cellular DNA, is the best evidence that the G-quadruplex exists in cells and is not just a test-tube oddity.
Once formed, the G-quadruplex and i-motif in the NHE III1 are relatively stable structures; indeed, the melting point of the G-quadruplex under physiological conditions is greater than 95°C. NM23-H2 is a member of a class of proteins that are implicated in transcriptional activation33 and were originally identified as nucleotide diphosphate kinases34; later, NM23-H1 was identified as a metastasis inhibitor.35 The NM23 proteins are found in prokaryotic and eukaryotic organisms, and the eukaryotic enzymes are a dimer of trimers with six active sites.34,36 In 1993, NM23-H2 was demonstrated to bind to the NHE III1 and activate transcription of MYC,37–39 but its precise role in this process remained controversial until recently. Confirmation of the role of NM23-H2 in transcriptional activation of MYC was obtained through siRNA experiments, and a correlation was found between NM23-H2 levels and MYC transcription.40 The enzyme was found to bind to the purine and pyrimidine single strands but not to the duplex, and mutation at the phosphate-binding site of NM23-H2 prevented binding of the purine- and pyrimidine-rich oligomers derived from NHE III1.40 Some of these results were independently confirmed in the Chowdhury laboratory.41 Molecular modeling of the purine-rich strand in an electropositive channel in one of the trimers (see inset in Figure 3B)40 demonstrated that three of the guanines could bind in the nucleotide phosphate pockets, with the intervening purines showing base stacking. This led to the proposal that NM23-H2 takes advantage of the dynamic character of the G-quadruplex and i-motif so that the enzyme unfolds these structures by trapping out the oligomer in a stepwise manner (a → b in Figure 3B).40 The requirement for a 44-mer suggests that the flanking single strand or duplex is bound at the equator region between the two trimers. Finally, the relatively low binding affinity of the NM23-H2 for the single-stranded purine- and pyrimidine-rich oligomers ensures a facile hand-off to single-stranded DNA-binding proteins such as CNBP and hnRNP K (b → c in Figure 3B).
Deregulated MYC is generally not due to an activating (or other) mutation, but it can be attributed to gene amplification, translocations, altered ploidy, or enhanced transcription.42,43 Increased MYC expression is usually sufficient to oncogenically transform cells in vitro and promote tumorigenesis in vivo.42,44–47 Its overexpression has been noted in a multitude of non-Hodgkin lymphomas, ranging from 10% to 100% of the patient population.15 Translocations involving the MYC locus define the rare but aggressive Burkitt lymphoma, and MYC mRNA overexpression correlates with inferior survival in the more prevalent and diffuse large B-cell lymphoma.48–50 To list and detail all the solid malignancies with which MYC aberrancies are associated would take a review of its own, but it is important to note that upregulation has been reported in up to 80% of all solid malignancies, including osteosarcomas, medulloblastomas, and skin, gastrointestinal, ovarian, and breast cancers.15,45,46,51–54 However, MYC is not a negative prognostic factor in all cancers. In estrogen receptor–positive breast cancer, MYC expression is critical to the response to tamoxifen (hormone therapy).55 Patients with colon cancer that have amplifications in MYC have better response rates to 5-fluorouricil, with increased progression-free and overall survival rates.56 When contemplating anti-MYC therapies for oncology, one must take into account the genetic profile and the type of disease, for not all cancers are created equal, especially with MYC.
As one of the central players in oncogenesis and a cooperative oncogene, altered MYC expression is often an early step in the multistage transformation and one on which all other mutations are based. Cancer cells appear to be addicted to a deregulated MYC, as proposed in 2008 by Weinstein.57 This addiction can be the Achilles’ heel of many cancers, offering the potential for a selective window for cancer versus normal cells. Recent work by two groups has highlighted the potential of MYC-targeted therapies.58,59 In both cases, MYC knockout was inducible and examined in a whole animal model or cell, both in vitro and in vivo. Soucek and coworkers used an inducible dominant negative knockout protein, termed Omomyc, to abrogate MYC activity. Their work demonstrated the key function of MYC, in cooperation with kRAS mutations, in the early transformation of lung cancer and the maintenance of established cancers. Neoplasias that were allowed to develop (as driven by overexpressed MYC and mutated kRAS) over 18 weeks were treated with Omomyc. After four weeks, no masses were evident, and only scattered atypical foci were noted. Perhaps the greater value of the study was the evaluation of systemic effects of MYC knockout. The researchers were able to show no overall detriment to health, although there was histologic evidence of damage to the rapidly proliferating tissues: skin, intestine, and testes. All these effects, however, were entirely reversible within a short period. Normal histology was recovered in less than seven days of Omomyc treatment cessation. This study demonstrated that whereas knocking out MYC can affect the normal tissues, the recovery rate is rapid and the tumors completely regress—and, in this case, never regrew.59 The authors were able to provide the first concrete evidence that molecularly targeted therapies specific to MYC may have a wide therapeutic window with both a specific and an efficacious antitumor effect.
Not all targeted therapies are able to achieve the 100% knockout used by Soucek and coworkers, and Dean Felsher’s laboratory has demonstrated that it is not necessary. Also using an inducible knockdown system (this time in T cells), they were able to titrate the repression of MYC and examine the minimal amount of reduction necessary for an antitumor response. In vitro MYC downregulation, but not ablation, was sufficient to significantly decrease cell density and growth, convert a proliferative state into an apoptotic one, and dramatically decrease cell number and cell size. In the whole animal studies, similar findings highlighted an apparent threshold level of MYC reduction (approximately 50%) required to block tumor growth or even regress it.58 It is clear that with MYC’s pivotal involvement in proliferation and the oncogenic process, it is a critical new target. And the recent evidence described demonstrates high potential for a large therapeutic window, especially in light of a threshold decrease of MYC beyond which oncogenesis cannot be restored.
As a direct consequence of understanding the dynamic character of the interplay among (1) transcriptionally induced negative superhelicity; (2) the equilibrium in the NHE III1 of duplex, single-stranded, and paranemic DNA structures (G-quadruplex and i-motif); and (3) the transcriptional factors that control this process (NM23-H2, nucleolin, hnRNP K, CNBP, and Sp1; Figure 3A), drug targeting of MYC transcription via stabilization of DNA secondary structures becomes an attractive possibility. Indeed, even before these details were known, it was correctly postulated that stabilization of the G-quadruplex in the NHE III1 would lead to repression of MYC.14 However, with a clear role for NM23-H240 and nucleolin32 in the unfolding or folding of the G-quadruplex (Figure 3A and and3B),3B), more strategies are possible involving specific inhibition of NM23-H2 or relocalization of nucleolin into the nucleus. Furthermore, the importance of negative superhelicity in controlling the processes suggests a potential auxiliary function of topoisomerase inhibitors in modulating the role of strategies that target the processes controlled by NM23-H2 and nucleolin. Also, the association of PARP with G-quadruplexes60 suggests that PARP inhibitors might play a role. In the following subsections, we describe experiments in which (1) the role of NM23-H2 in activating MYC transcription is modulated by G-quadruplex-interactive compounds40 and (2) relocalization of nucleolin from the nucleolus into the nucleus is proposed to inhibit MYC transcription via the G-quadruplex in the NHE III1.
Because NM23-H2 takes advantage of the dynamic character of the G-quadruplex and i-motif to catalyze the stepwise unfolding of these structures (Figure 3B), ionic conditions or compounds that stabilize either structure should inhibit this process.40 Indeed, KCl, which stabilizes the G-quadruplex, showed a 90% inhibition of NM23-H2-catalyzed unfolding of the G-quadruplex at 100 mM, as compared to no KCl.40 Likewise, in a dose-dependent manner, a 10-fold increase in concentration of TMPyP4 (1–10 µM) showed a dramatic decrease in NM23-H2 unfolding of the G-quadruplex. In comparison, TMPyP2, a positional isomer of TMPyP4, binds weakly to the MYC G-quadruplex and does not alter NM23-H2 binding so dramatically.40
On the basis of the preferential stabilization and associated inhibition of NM23-H2-catalyzed unfolding of the G-quadruplex by TMPyP4 (d → b in Figure 3B), TMPyP4 should preferentially inhibit MYC transcription. To evaluate the comparative effect of TMPyP4 and TMPyP2 on transcription from the MYC promoter, a luciferase reporter assay was used with either the wild-type or a specific mutant sequence that destabilized the G-quadruplex, as well as with two Burkitt lymphoma cell lines in which one allele had undergone a translocation between chromosomes 8 and 14. In the latter case, the Ramos cell line retained the NHE III1, which was under the control of the heavy chain immunoglobulin region, whereas the CA46 lost the NHE III1 during the translocation and, thus, the primary target for modulation of gene transcription (Figure 2B). In both assays, the results were as anticipated. TMPyP4 targeting of the wild-type sequence has a dramatic effect on luciferase expression, and targeting the mutant sequence with TMPyP4 partially “rescued” the destabilized G-quadruplex sequence.14 With the Burkitt lymphoma cell lines, there was a significant effect on MYC expression in the only case of the Ramos cell line, which retained the target for TMPyP4 in the NHE III1 region of DNA (Figure 2C). TMPyP2 had only marginal effects on the Ramos cell line, and neither compound had any effect in the CA46 cell line (Figure 2C). The use of the Burkitt lymphoma cell lines in demonstrating the selective effect on MYC has been independently verified by other groups using structurally distinct G-quadruplex-interactive molecules, including quindolines61 and actinomycin D.62
Quarfloxin (CX-3543) is a fluoroquinolone-based antitumor agent derived from norfloxin via A-62176 and QQ58 (Figure 4A). The latter compound has a mixed mechanism of action as a topoisomerase II poison and a G-quadruplex-interactive compound.63 Medicinal chemistry at Cyternex (now Cylene, San Diego, CA) significantly increased the selectivity for G-quadruplex over duplex DNA, which eliminated the topoisomerase II poisoning effect. Although CX-3543 was selected for its interaction with the MYC G-quadruplex, it was found to be preferentially concentrated in the nucleoli of cancer cells where it inhibits pol I.64 The selective inhibition of pol I is likely due to the binding of CX-3543 to G-quadruplexes in the rDNA, where a search algorithm had identified 14 putative G-quadruplex-forming sequences on the nontemplate strand. CX-3543 was found to disrupt 13 of the 14 nucleolin/G-quadruplex complexes.64 The result of disruption of these complexes is a redistribution of the nucleolin into the nucleoplasm.64 This redistribution of nucleolin mimics a stress response that can be caused by cytotoxic agents, heat shock, or gamma irradiation. This leads to apoptosis and cell death in cancer cells, which is attributed to a number of mechanisms. For example, complex formation between nucleolin and replication protein A in stressed cells65,66 leads to inhibition of DNA replication following heat shock and gamma irradiation. In addition, stress conditions that lead to nucleolin mobilization induce nucleolin:p53 complex formation.67 Likewise, p53 ubiquitination by Hdm2 can be prevented by binding of nucleolin to Hdm2. The overall effect is activation and stabilization of p53. We have shown that nucleolin binds to the MYC G-quadruplex to suppress transcription,32 but this seems at odds with the observation that MYC inhibition was not observed in the published CX-3543 work.64 However, the authors examined the effect of CX-3543 on MYC after only two hours, which is likely too short to see the downstream effect on transcription via disruption of the nucleolin/rDNA complex and displacement of nucleolin to the MYC G-quadruplex. In contrast to the finding of this report, MYC transcription was found to be inhibited by 85% in tumor tissues taken from mice bearing HCT-116 colorectal tumors after treatment with CX-3543.68 Quarfloxin is presently in phase 2 clinical trials for neuroendocrine/carcinoid tumors and is a first-in-class G-quadruplex-interactive drug targeted at parallel-stranded G-quadruplexes typified by that found in the MYC promoter.
MYC is not the only oncogene or gene promoter region to possess a G-quadruplex or i-motif. Bioinformatics studies have demonstrated that a large number of transcriptional and translational promoter regions have putative quadruplex-forming sequences in proximity to the transcription start site.69–71 Indeed, there is an abnormally high concentration of these elements within 500 base pairs on each side of the transcriptional start site, in an overrepresentation of genes associated with cell signaling.71 Studies from this lab and others have shown that these structures are found in the promoters of Bcl-272, VEGF,73,74 Hif-1α,75 PDGFA,76 PDGFR-β,77 MYB,78 hTERT,79 c-Kit,80–82 kRAS,83 and Rb,84 and this list grows annually. Moreover, promoter regions from at least one oncogene within each of the six hallmarks of cancer—self-sufficiency in growth signals, insensitivity to anti-growth signals, apoptotic evasion, angiogenesis, limitless potential for replication, and tissue invasion and metastasis—contain G-quadruplexes with some involvement in transcriptional control.15,85
One concern with the prevalence of G-quadruplex structures within a multitude of promoters has been achieving target specificity with a small molecule drug. In this respect, the problem is similar to that anticipated with kinase inhibitors in the early days: a similar binding pocket spread over a variety of molecular targets. However, the G-quadruplex structures do have a few advantages. These secondary DNA structures are globular and more analogous to tertiary protein structures than to repetitive, linear, double-stranded B-DNA. Additionally, the G-quadruplex structures vary in the number of tetrad stacks and in the lengths of, and bases composing, the loops connecting the tetrads.86 The MYC G-quadruplex is an all-parallel structure made of three stacking tetrads and connecting loops of A:TG:A (5′–3′) making a 1:2:1 loop isomer (see Figure 1). Both VEGF and HIF-1α are similar parallel structures with three stacked tetrads but have 1:4:1 and 1:6:1 loop isomers with differing bases in the loops.73–75 Structures also range from these straightforward G-quadruplexes to the more complex formation within the hTERT promoter, the region of which is proposed to form two tandem G-quadruplexes connected by six or seven bases. The 5′ G-quadruplex is an all-parallel formation with loop lengths of 1:3:1, whereas the 3′ G-quadruplex is a unique structure with mixed-parallel and antiparallel loops with lengths of 3:26:1. Moreover, the 26-base middle loop forms its own secondary structure, a hairpin. Class specificity is easy to visualize, and these loop differences within similar structures potentially allow for specificity in drug targeting.
MYC deregulation is a pivotal turning point in the formation of many liquid and solid cancers to which the tumor cells are apparently addicted.15,57 This, in combination with a couple of key publications validating the potentially wide therapeutic window,58,59 has led to a number of drug discovery programs focused on MYC-targeted therapies for protein inhibition. To date, no successful small molecule has been developed to specifically and potently downregulate, or inhibit the activity of, MYC expression, although based on the experience with kinase inhibitors, it may not be necessary or even desirable to achieve high selectivity. Insight into the mechanisms of transcriptional control within the NHE III1 region of the promoter via the formation and resolution of unique DNA secondary structures has provided a new molecular target and strategy to achieve external modulation of MYC expression with small drug-like molecules.
Within the primary region of transcriptional control of the MYC gene lies a dynamic GC-rich region of DNA that is capable of forming G-quadruplexes and i-motifs, in addition to the single- and double-stranded forms. The non-B-DNA forms serve as the silencing elements and on/off switches for MYC expression. As such, they are great potential targets for therapeutic development (as reviewed in reference 87). In fact, a number of groups around the world are currently pursuing this mechanism of action for new compounds. However, although we believe that the accumulation of increasing supporting data for the existence of promoter G-quadruplex structures and their involvement in transcriptional control is compelling, some skeptics still believe this to be another story of a test-tube oddity without a true biological mechanism or formation in cells.
Some of the best evidence for the existence of unique DNA topologies in vivo is the identification of proteins that bind to or control their formation, which is still largely lacking with Z-DNA.1 Within the NHE III1 region of the MYC promoter, the transcriptional factors Sp1 and CNBP and hnRNP K bind to the double- and single-stranded forms of DNA, respectively, whereas nucleolin and NM23-H2 assist in the formation and resolution, respectively, of the G-quadruplex structures.9–12,32,40 The recent identification of nucleolin and NM23-H2 as critical players in MYC G-quadruplex modulation, both ex vivo and in cells, serves as a strong validation.32,40 Cationic porphyrins that alter protein binding to the NHE III1 ex vivo then decrease MYC expression in cells, further substantiating the existence of G-quadruplexes in cells and euchromatin.14,40
The ultimate proof for the existence of these structures in cells will come when pulldown experiments similar to those using specific antibodies to transcriptional factors demonstrate that these structures are located in the elements that have putative G-quadruplex-forming sequences. ChIP experiments showing nucleolin binding to the NHE III1 in the MYC promoter are the closest thing that we have to proof for the existence of G-quadruplexes in promoter regions.32 In contrast to just a few years ago, there is considerable worldwide activity in this, as yet, still little-known area. So far, the story continues to hold water, and there may be schools of fish yet to be harvested in this exciting new area for molecular targeted therapy.
We are grateful to David Bishop for preparing, proofreading, and editing the final versions of the article and figures.
We have discovered, after this paper was accepted for publication, that Quarfloxin has now been withdrawn from clinical studies for neuroendocrine tumors.
The authors’ research is supported by grants from the National Institutes of Health (CA95060 and GM085585), the National Foundation for Cancer Research (VONHOFF0601), and the Leukemia & Lymphoma Society (6225-08).
Laurence Hurley is a shareholder in Cylene Pharmaceuticals. Tracy A. Brooks declares no conflicts of interest with respect to the publication of this article.