Aiming at Neuroblastoma and Hitting Other Worthy Targets

Abstract

Neuroblastoma is, at once, the most common and deadly extracranial solid tumor of childhood. Efforts aimed at targeting the neural characteristics of these tumors have taught us much about neural crest cell biology, apoptosis induction in the nervous system, and neurotrophin receptor signaling and intracellular processing. But neuroblastoma remains a formidable enemy to the oncologist and an enigmatic target to the neuroscientist.

Neuroblastoma is a cancer that occurs almost exclusively in children and that derives from the sympathetic nervous system, usually in the chest or abdomen. The origin of neuroblastoma cells is the primitive neural crest and as such, its histological appearances include cells with neuronlike, Schwann cell–like, and stem cell characteristics. These cells variably express neurotransmitters, neurotransmitter receptors and uptake proteins, neurotrophin receptors, and Schwann cell proteins.¹ Thus, targeting therapies to neuroblastoma might best involve targeting these classes of molecules. Not surprisingly, this is easier said than done. But the paths traversed in the quest for therapies for neuroblastoma have enriched our understanding of neurotrophin receptors and glutathione synthetic enzymes, changed our understanding of anti-apoptotic signaling, and informed our view of Alzheimer disease as a disorder of late development and protein processing.^2,3

Off-Target Effects? Find a New Drug or a New Target

One of the first approaches we took to neuroblastoma was to exploit the catecholamine uptake system on the surfaces of many neuroblastoma cells by administering an oxygen radical–generating catecholamine, 6-hydroxydopamine.⁴ The notion was that neuroblastoma cells would actively concentrate this drug, which would, in turn, generate toxic reactive oxygen species in situ and destroy these tumor cells. Others had administered 6-hydroxydopamine intrathecally to destroy central dopaminergic neurons in an early model of Parkinson’s disease.⁵ Studies that administered 6-hydroxydopamine parenterally to mice demonstrated destruction of the sympathetic nervous system. In fact, destruction of the sympathetic nervous system by 6-hydroxydopmaine impaired the growth of subsequently implanted subcutaneous neuroblastomas.⁶

Given these prior studies, it was not surprising that our initial studies of 6-hydroxydopamine alone demonstrated that doses high enough to ablate neuroblastomas also were prohibitively toxic to the sympathetic nervous system. So we set about identifying a compound that would selectively protect normal cells while leaving the neuroblastoma cells susceptible to oxygen radical attack. We thought it might be possible to find such a compound because of our experience many years earlier with a compound that had been developed at the Walter Reed Army Institute of Research as a radioprotective agent.⁷ We had piloted its use as a mucolytic agent in patients with cystic fibrosis.⁸ If ever there was an example of the nonlinear path to discovery, this was it!

Amifostine (WR2721; ethiofos) is a thiophosphate that is actively taken up and cleaved to a sulfhydryl by most normal cells but not by most cancer cells.⁷ The sulfhydryl compound reduces reactive oxygen species, itself getting oxidized to a disulfide in the process. We reasoned that adjunctive use of amifostine with 6-hydroxydopamine might afford protection to normal cells without compromising toxicity to neuroblastoma cells. However, although single-dose studies in mice looked promising, repeated doses produced synergistic toxicity to normal cells.^4,9 As is illustrated in Figure 1, it turned out that the disulfide metabolite of amifostine, which was assumed to freely cross the cell membranes of both normal and cancer cells, is among the most potent inhibitors of γ-glutamylcysteine synthetase—the rate-limiting enzyme for glutathione synthesis—known.¹⁰ Glutathione, in turn, is the one of the most important intrinsic antioxidants; inhibiting its synthesis with a metabolite of a putative extrinsic antioxidant is a curious irony, indeed! These studies characterized both a novel reagent for biochemical studies of glutathione synthesis and a potential mechanism for the cumulative side effects of amifostine.¹¹

Figure 1.

Amifostine (R-S-OPO3) is preferentially cleaved to a sulfhydryl (R-SH) in normal cells relative to tumor cells. In these normal cells, R-SH acts as an antioxidant, reducing reactive oxygen species and, in turn, becoming oxidized (to R-S-S-R) itself. R-S-S-R reacts with γ-glutamylcysteine synthetase, the rate-limiting enzyme for glutathione synthesis, and glutathione synthesis ceases.

Discovery of the unfortunate “fit” and covalent affixing of the disulfide of amifostine into the active site of γ-glutamylcysteine synthetase led to structure-activity relationship studies aimed at identifying a normal-selective protective agent, the disulfide of which would not interfere with glutathione synthesis.¹² This work provided some of the early stereochemistry of the active site of this important enzyme and raised the possibility of dissociating the efficacy from the toxicity of cell-selective antioxidants, a concept under most intensive study in the area of neuroprotection, rather than cancer.

“When You Come to a Fork in the Road, Take It.”—Yogi Berra

A central dogma of biology as it was taught 2 or 3 decades ago was “One gene, one polypeptide.”¹³ The notion that prevailed at the beginning of the molecular genetic era was that if one knew the gene sequence, one knew the one and only one protein that was made off of it. We now know that this is far from the truth—that many more than one protein can be made off of a single genetic template, that proteins frequently get made from noncontiguous genetic sequences, that some of the most “important” genes do not make proteins at all.^14–16 Similarly, the dogma with respect to receptor-ligand pairs was that if one knew the receptor-ligand pair, one could unambiguously predict the array of signal transduction pathways that would be activated in a given type of cell.¹⁷ We now know that the same receptor-ligand pair in a single cell type can initiate different downstream pathways in different circumstances and milieu and that proximal signaling pathways can branch further downstream and result in diametrically opposed outcomes for the cell depending on the branch that is enacted.^18,19

Why is this important for our patients? As an example, consider the role of the TrkA receptor and its ligand, nerve growth factor (NGF) in a developing sensory or sympathetic neuron. Mutations in the intracellular tyrosine kinase domain of TrkA result in the syndrome of congenital insensitivity to pain and anhydrosis.²⁰ A normal intracellular domain of TrkA is therefore critical for the normal development and function of sensory and sympathetic neurons.²⁰ On the other hand, consider the role of the TrkA receptor and NGF in a pheochromocytoma cell. The signal transduction pathway triggered by binding of NGF to TrkA depends on the level of expression of TrkA. Native-level expression of TrkA leads to MEK1-ERK1/2-p-CREB-mediated antiapoptotic signaling and cell survival; overexpression of TrkA in the same cells leads to MEK3/6-p-P38MAP-mediated proapoptotic signaling and cell death.²¹ Therefore, under some circumstances and with regard to some cellular contexts, antagonists of NGF signaling through TrkA would be therapeutic, and under others, it would be pathogenic.

Another example of a branch point distal to a receptor-ligand pair involves another neurotrophin receptor, p75NTR. p75NTR was originally thought only to be a coreceptor that bound to TrkA and, in so doing, enhanced the affinity of TrkA for NGF. It is now known that, in addition to its coreceptor function, p75NTR is an independently signaling receptor with a so-called death domain sequence that led to its identification as a proapoptotic protein.¹⁹ However, dependent on cellular and environmental context, but not obviously related to level of expression, p75NTR can signal through phosphoinositol-3-kinase and Akt to be antiapoptotic²² or through de novo cholesterol synthesis to be proapoptotic.^23,24 Although branch points from a receptor-ligand pair can complicate using the receptor or the ligand as a therapeutic target, knowing the alternative downstream effectors of receptor-ligand signaling identifies novel targets that can be unique to a particular outcome of receptor-ligand binding.

Developing targeted therapies for our patients that truly target species responsible for either the pathologic outcome or its desirable counterpart requires an understanding of the branch points and contingencies inherent in the complex network we now know signaling to be. Understanding these pathways can predict otherwise unforeseen therapeutic failures or side effects and can identify more specific targets for effective, nontoxic therapies.²⁵

Location, Location, Location

If we were going to solve the problem of chemotherapeutic resistance of neuroblastoma cells, given the dichotomous behavior of their neurotrophin receptors vis-à-vis life and death of the cell, we needed to evolve an algorithm to predict whether expression of a particular neurotrophin in a particular neuroblastoma would be proapoptotic or antiapoptotic. We knew that translating such an algorithm from the single cell into the cell-heterogeneous tumor would be an enormous challenge, but at least understanding the mechanism at the life-or-death decision point would be a start.^26–28 The story ends up being more complex and more interesting than we could ever have imagined.

Our initial studies used hydrogen peroxide or 6-hydroxydopamine as the in vitro “chemotherapy” and asked by what mechanism p75NTR was proapoptotic or antiapoptotic in neural crest tumor cells. In PC12 rat pheochromocytoma cells, p75NTR was antiapoptotic and functioned as a cytoplasmic antioxidant, enhancing recycling of glutathione to its reduced form. The antiapoptotic, antioxidant effects of p75NTR could be replicated with the intracellular domain of p75NTR alone; that is, intracellular expression of the intracellular domain of p75NTR in cells that did not make holo-p75NTR resulted in protection from hydrogen peroxide-induced glutathione oxidation and apoptosis.²⁶

In SH-EP1 human neuroblastoma cells, p75NTR was proapoptotic by itself, but facilitated the antiapoptotic activity of TrkA when they were engineered to coexpress TrkA and p75NTR with a p75NTR/TrkA ratio less than 10/1.²⁷ Perhaps knowing the ratio of p75NTR/TrkA in the cell would allow us to predict whether the preponderant effect of NGF binding to its receptors would be proapoptotic or antiapoptotic.

But neither hydrogen peroxide nor 6-hydroxydopamine was used clinically in neuroblastoma. We were anxious to discern the potential role of p75NTR in chemoresistance of clinical neuroblastoma. We treated SH-EP1 cells with the chemotherapeutic agent fenretinide, an agent that is in clinical trials specifically for neuroblastoma.²⁸ Fenretinide is hypothesized to have several chemotherapeutic mechanisms, one of which is the generation of reactive oxygen species.²⁹ We and others^29,30 have demonstrated that fenretinide specifically induces accumulation of mitochondrial superoxide.

Expression of p75NTR or just its intracellular domain enhances induction of apoptosis by fenretinide in SH-EP1 cells. In fact, it enhances accumulation of mitochondrial superoxide after fenretinide treatment by a mechanism that can be inhibited by inhibition of mitochondrial complex II or by pretreatment with mitochondria-specific, but not cytoplasm-specific, antioxidants.³⁰ The effectiveness of p75NTR and its intracellular domain in the mitochondrion led us to examine the mechanism by which p75NTR works.

As an independently signaling receptor,²⁷ p75NTR forms an asymmetric dimer in the cell membrane and binds to NGF or pro-NGF. This complex recruits one or more interactors that bind to the intracellular domain.¹⁹ The intracellular domain is then cleaved from the holo-receptor by α- and γ-secretase and is thought to be a transcription factor.^31,32 In addition, there is recent evidence that the intracellular domain of p75NTR goes directly to the mitochondria.³³ This latter finding makes it tempting to hypothesize that the mitochondrial antioxidant effect is a direct effect of the intracellular domain of p75NTR. But given that mitochondrial complex II, inhibition of which has a mitochondrial antioxidant effect, is composed of 4 subunits all of which are encoded by nuclear genes,³⁴ it is possible that the mitochondrial antioxidant effect of p75NTR intracellular domain is the result of its effect on transcription of the subunit proteins of complex II. This possibility is currently under investigation.

Why are we working so hard to understand the mechanisms behind the effects of p75NTR on the cell? If expression of p75NTR by neuroblastoma cells enhances the effectiveness of fenretinide against them, why does it matter how this occurs? For one thing, expression of p75NTR by neuroblastoma cells likely varies, not only from neuroblastoma to neuroblastoma, but from cell to cell within a single neuroblastoma. But if we knew the downstream pathway that resulted in its effects, we could perhaps mimic this activity with molecules smaller than proteins. For another, understanding the mechanisms of action of a developmentally- and spatially-regulated receptor in the central and peripheral nervous system might lead to unexpected revelations about nervous system disorders other than neuroblastoma. An evolving example, depicted in Figure 2, follows.

Figure 2.

p75NTR is a trans-membrane protein. Cleavage of p75NTR by α- and γ-secretase (labelled “γ”) liberates its intracellular domain (p75ICD). p75ICD is thought to translocate to the nucleus and mitochondria, altering its transcriptome and redox state.

Aging is Development for Grown-ups

In the embryonic human brain, p75NTR is ubiquitously expressed. As the brain develops, however, its expression is increasingly restricted. In the adult brain, p75NTR is most abundant in the basal forebrain (particularly in the nucleus basalis of Meynert), the hippocampus, and the cerebellum.³⁵

In the course of our studies of p75NTR, we decided to study differences in the transcriptome between PC12 cell lines that either did or did not express p75NTR in their native state. As it happens, we were guided in performance of these microarray studies and interpretation of their results by a postdoctoral fellow in the laboratory next door to ours. She was working on Alzheimer disease and comparing the whole-brain transcriptomes of mice that expressed wild-type or familial Alzheimer disease mutant presenilin, respectively.³⁶ At the end of the side-by-side experiments, we (with our difference map of p75NTR-positive vs p75NTR-negative PC12 cells) and she (with her difference map of wild-type presenilin vs mutant presenilin brains) sat down to discuss the next step—data mining and interpretation. To all of our shock, the differences were almost superimposable! The same set of mRNAs was altered in the 2 very different scenarios. We came to the conclusion that p75NTR and presenilin must be part of the same signaling pathway.³

This should not have been too surprising to us. p75NTR is activated by cleavage by γ-secretase and presenilin is a member of the γ-secretase family of proteins.³⁷ Furthermore, the brain loci most affected in Alzheimer disease are among those in which expression of p75NTR persists throughout life.²⁴ But it was not until this serendipitous juxtaposition of 2 seemingly unrelated datasets that we began thinking of Alzheimer disease as a developmental aberration, the seeds of which could have been sown or at least prepared at some remote time in the past.³⁸

Perhaps most interesting from a clinical standpoint, in both cases, expression of the 5 major enzymes involved in cholesterol biosynthesis is altered. In fact, expression of 7-dehydrocholesterol reductase, 3-hydroxy-3-methyl-glutaryl–coenzyme A (HMG-CoA) reductase, diphospho-mevalonate decarboxylase, geranyl-trans transferase component B, and farnesyl diphosphate synthase is coregulated with that of p75NTR.²⁴ In addition, treatment of oxidant stress-resistant p75NTR-positive PC12 cells with mevastatin, an HMG-CoA reductase inhibitor, converts their sensitivity to oxidant stress to that of p75NTR-negative cells,²³ implying that the oxidant resistance conferred by expression of p75NTR is dependent on the activity of HMG-CoA reductase.

It is not yet clear what all of this will mean for attempts to stem the tide of Alzheimer disease using statins or γ-secretase inhibitors, particularly as patients with sporadic Alzheimer disease do not express mutant presenilin³⁹ and as it is not clear whether p75NTR is a pro-oxidant or antioxidant in neurons of the basal forebrain, hippocampus, or cerebellum.¹⁹

But it is clear that exploiting these findings clinically will require development of more specific drugs that target specific γ-secretases or that affect specific branch points off of the cholesterol biosynthetic pathway.²⁵

Less Is Sometimes More

The 26-kDa protein Bcl-2 is a mitochondrial protein that prevents apoptosis.⁴⁰ In neural crest cells, including cells of the neural crest tumor, neuroblastoma, overexpression of Bcl-2 family members results in enhanced sulfhydryl content of the cell.⁴¹ That is, recycling of oxidized (ie, disulfide) cytoplasmic glutathione to its reduced (ie, sulfhydryl) state is more efficient in cells that overexpress Bcl-2.⁴² This was intriguing to us because, although Bcl-2 makes cancer cells resistant to conventional chemotherapeutic agents, it might be expected to enhance the sensitivity of cancer cells to chemotherapeutic prodrugs that are activated by sulfhydryl reduction.⁴³ One such prodrug is the antimitotic, proapoptotic natural product, neocarzinostatin.⁴⁴

Sulfhydryl activation of neocarzinostatin results in formation of a DNA-cleaving agent that induces mitotic arrest, differentiation, and subsequent apoptosis in neuroblastoma cells in culture.⁴⁴ Overexpression of Bcl-2 indeed potentiates this effect.⁴³ We assumed that this was because neocarzinostatin would be activated to a greater extent in the more reducing environment of Bcl-2-overexpressing cells than in native cells. But this mechanism turned out to be unlikely.

Shortly after we published our work on the potentiation of the activity of neocarzinostatin by Bcl-2, data were published that made it clear that Bcl-2 exerted its conventional activity downstream of neocarzinostatin activation.⁴⁵ Even if more reduced glutathione resulted in more activated neocarzinostatin, Bcl-2 should have prevented that activated neocarzinostatin from inducing apoptosis. How, then, was Bcl-2 enhancing apoptosis induction by neocarzinostatin?

One clue came when we asked whether other cancer cell lines transfected with a Bcl-2 expression construct would exhibit the same potentiation of neocarzinostatin as our neuroblastoma cells. One breast cancer cell line, MCF-7, did not demonstrate this phenomenon. MCF-7 cells differed from the other cell lines we tested in that they did not express the enzyme caspase-3.⁴⁶ Caspase-3 cleaves Bcl-2 from a 26-kDa antiapoptotic protein to a 19-kDa proapoptotic protein.⁴⁷

Treatment of neuroblastoma cells with neocarzinostatin resulted in cleavage of Bcl-2 to its 19-kDa fragment.⁴⁸ Pretreatment of neuroblastoma cells with a caspase-3 inhibitor prevented this effect. Treatment of neuroblastoma cells with cisplatin and treatment of MCF-7 cells with neocarzinostatin did not result in cleavage of Bcl-2. But transfection of MCF-7 cells with a caspase-3 expression construct followed by treatment with neocarzinostatin resulted in cleavage of Bcl-2.⁴⁶ In accordance with this, transfection of neuroblastoma cells with Bcl-2 resulted in potentiation of neocarzinostatin-induced apoptosis, whereas transfection of MCF-7 breast cancer cells with Bcl-2 did not. However, transfection of caspase-3-transfected MCF-7 cells with Bcl-2 resulted in potentiation of neocarzinostatin-induced apoptosis.⁴⁶ Thus, both Bcl-2 and caspase-3 were necessary and sufficient for potentiation of neocarzinostatin-induced apoptosis. What is more, extending this evaluation to many tumor types and measuring Bcl-2 and caspase-3 content in each case afforded an algorithm by which one could predict susceptibility of any given cell to neocarzinostatin-induced apoptosis.⁴⁹

But neocarzinostatin is expensive to isolate and purify, immunogenic in humans, and light- and heat-labile when stored on the shelf.⁵⁰ It is not the perfect chemotherapeutic agent.

The Value of the Near or Not-So-Near Miss

New techniques that look in an unbiased fashion at all of the transcription products made in a given cell or tumor are already identifying novel targets unique to particular subsets of tumor cells in the neural crest lineage. For example, Schwannomas express neither the neural markers doublecortin and Kidins220 nor the cytoskeletal protein strathmin-like 2. Ganglioneuromas express small amounts of all 3. Ganglioneuroblastomas express yet higher amounts and neuroblastomas express the highest amounts of these proteins. Thus, the expression profile of a given neural crest tumor allows determination of its degree of differentiation and, conversely, its degree of malignancy.⁵¹ Studies of this kind can identify targets for chemotherapy specific to a particular patient’s tumor. Similarly, studies of the mitochondrial electron transport chain have recently revealed that release of reactive oxygen species leading to cell death, previously ascribed to complexes I and III, can also come from complex II.⁵² Complex II is therefore a drugable target for cancer chemotherapy. This appears to be at least one of the mechanisms of action of fenretinide, a chemotherapeutic agent in phase II studies for neuroblastoma.^28–30

The evolution of these studies from looking for the chemotherapeutic panacea to looking for the characteristics of the tumor and host that would predict effectiveness of a particular chemotherapeutic regimen serves as a prototype for the future of cancer therapeutics—individualization of therapy based on the tumor, the host, the environment, and the interaction among the 3. These studies taught us the importance of understanding the enemy before designing the magic bullet that takes it down. Context, it seems, is everything.