Mechanisms of neuroblastoma regression

Abstract

Recent genomic and biological studies of neuroblastoma have shed light on the dramatic heterogeneity in the clinical behaviour of this disease, which spans from spontaneous regression or differentiation in some patients, to relentless disease progression in others, despite intensive multimodality therapy. This evidence also suggests several possible mechanisms to explain the phenomena of spontaneous regression in neuroblastomas, including neurotrophin deprivation, humoral or cellular immunity, loss of telomerase activity and alterations in epigenetic regulation. A better understanding of the mechanisms of spontaneous regression might help to identify optimal therapeutic approaches for patients with these tumours. Currently, the most druggable mechanism is the delayed activation of developmentally programmed cell death regulated by the tropomyosin receptor kinase A pathway. Indeed, targeted therapy aimed at inhibiting neurotrophin receptors might be used in lieu of conventional chemotherapy or radiation in infants with biologically favourable tumours that require treatment. Alternative approaches consist of breaking immune tolerance to tumour antigens or activating neurotrophin receptor pathways to induce neuronal differentiation. These approaches are likely to be most effective against biologically favourable tumours, but they might also provide insights into treatment of biologically unfavourable tumours. We describe the different mechanisms of spontaneous neuroblastoma regression and the consequent therapeutic approaches.

Introduction

Neuroblastoma is the most common extracranial solid tumour of children; it accounts for 8–10% of childhood cancers in the USA and Europe.^1–4 Neuroblastomas in children 18 months of age or older are frequently unresectable or metastatic, require intensive multimodality therapy and are associated with a 40–50% survival rate.^1,2,5 However, neuroblastomas in children under 18 months of age behave very differently. Most infants, even with metastatic disease, can be cured with moderate-intensity chemotherapy, and some patients with a special pattern of metastasis have a high likelihood of undergoing spontaneous regression without chemotherapy.^6–10 Indeed, the prevalence of spontaneous regression has been documented by mass-screening programmes undertaken in Japan, Quebec and Europe.^11–14 Furthermore, children (and adults) can present with localized, benign ganglioneuromas, which likely represent neuroblastic tumours that have become differentiated.^15–19 The exact mechanisms responsible for spontaneous regression (and differentiation) are uncertain, but several plausible mechanisms have been proposed to explain these phenomena.^6–10 In this Review, we explore the current understanding of the genomic, biological and immunological mechanisms that underlie spontaneous regression, and possible approaches to therapy.

Genetic predisposition

About 1–2% of patients with neuroblastoma have a family history of this disease.^20–23 Two genes have been identified, ALK and PHOX2B, that account for ~80% of hereditary neuroblastoma. Specifically, several groups have shown that activating mutations in ALK are responsible for ~75% cases of hereditary neuroblastoma.^20,24–26 Neuroblastomas also occur in patients with congenital central hypoventilation syndrome (Ondine’s curse), and inactivating mutations of PHOX2B are present in most of these patients, accounting for another 5% of hereditary cases.^22,27,28

Genome-wide association studies have identified several gene polymorphisms associated with a low, but significant risk of neuroblastoma, which include BARD1, LMO1, and LIN28B among others.^29–33 Genetically engineered mouse models that develop neuroblastoma are available, and include TH−MYCN,³⁴ MDM2+MYCN,³⁵ ALK+MYCN,³⁶ ALK,³⁷ and LIN28B.³⁸ However, no cases of hereditary neuroblastoma have been associated with germline mutations of genes other than ALK and PHOX2B, and none has been associated with spontaneous regression.

Subtypes of neuroblastomas

Several genomic alterations have been identified in neuroblastomas that generate tumours with distinct genotypes characterized by different patterns of clinical behaviour (Figure 1).¹⁵ Neuroblastomas can be divided into three major subtypes based on cytogenetic profiles—subtype 1, 2A and 2B. Subtype 1 is characterized by numerical chromosome alterations resulting in hyperdiploidy or near triploidy, but with few if any segmental chromosomal abnormalities (SCAs).^39,40 High expression levels of the tropomyosin receptor kinase (Trk) A is common to almost all subtype 1 tumours. Patients with these tumours have favourable features (such as young age and low tumour stage) and outcomes. Subtype 2 is characterized by near diploidy and recurrent SCAs. Many of these tumours have an unbalanced gain of the chromosome 17q, and most overexpress both TrkB and its ligand. However, this subtype can be further subdivided into subtype 2A, which frequently has segmental loss of chromosomes 3p, 4p, and/or 11q; and subtype 2B, which has deletion of chromosome 1p and/or MYCN amplification. Neuroblastoma of subtypes 2A and 2B are associated with older age, advanced tumour stage and a worse clinical outcome, with subtype 2B tumours being the most aggressive (Figure 1).¹⁵

Figure 1.

Genomic model of neuroblastoma development. The major genomic pathways and genotype subsets of neuroblastoma are depicted here. Type 1 neuroblastomas have a favourable clinical outcome and consistently show numerical chromosome abnormalities (near-triploidy) without SCAs. They also have high expression of the TrkA neurotrophin receptor, and they are prone to undergo spontaneous regression (or differentiation), depending on the presence (+) or absence (−) of NGF in their microenvironment, respectively. Conversely, type 2 neuroblastomas are unfavourable clinically and are characterized by SCAs. Many of these tumours have unbalanced gain of chromosome 17q and express TrkB and BDNF. They can be separated into two subtypes based on additional genomic changes: type 2A tumours also have selective regional loss of 3p, 4p, and/or 11q, and many also have gain of chromosome 7; and type 2B have MYCN amplification, usually with 1p deletion, and they generally lack the additional changes found in type 2A. Type 2B tumours are the most aggressive and rapidly progressive subtype. Abbreviations: BDNF, brain-derived neurotrophic factor; NGF, nerve growth factor; SCA, segmental chromosomal abnormalities.

Deep-sequencing studies of neuroblastoma exomes or whole-genome analysis have identified relatively few additional gene mutations that were not otherwise known to have a role in this disease. In addition to MYCN amplification (found in 22% of primary tumours), activating mutations or rearrangements of ALK were found in 8–10% of sporadic tumours.⁴¹ Furthermore, mutations in ATRX, ARID1A, ARID1B, MYCN, PTPN11 and NRAS were found in 1–3% of cases (Table 1).^15,41–44 Oncogenic activation of FOXR1 by 11q23 intrachromosomal deletion-fusions has been identified in a few neuroblastoma cases (Table 1).⁴⁵ In general, these alterations are associated with high-risk disease, and ATRX mutations are more common in older patients.⁴² However, with the exception of hyperdiploidy and/or near triploidy with numerical chromosome alterations, no specific genomic changes are associated with low-risk disease or with disease regression.^46–48 The relative paucity of mutations in neuroblastomas, especially in patients with low-risk tumours, suggests that copy number, gene dosage and/or epigenetic regulation are probably important mechanisms in regulating gene expression and tumour cell behaviour.

Table 1.

Somatically acquired mutations and rearrangements in neuroblastoma

Gene	Function	Alteration	Frequency (%)	Reference
MYCN	Transcription factor	Amplification Activating mutation	22 1.7	15 41
ALK	Tyrosine kinase receptor	Activating mutation Amplification/duplication	9.2 3.4	41 41
PTPN11	Tyrosine phosphatase	Activating mutation	2.9	41
ATRX	Chromatin remodelling	Inactivating mutation	2.5	42,44
ARID1A/B	Chromatin remodelling	Inactivating mutation	2–3	43
NRAS	Signalling protein	Activating mutation	~1	41
MLL-FOXR1, or PAFAH1B-FOXR1	Transcription factor	Translocation	1.3	45

Neuroblastoma—spontaneous regression

Historical descriptions

Spontaneous regression of cancer is defined as the decrease in size or disappearance of a primary tumour or metastatic disease without therapeutic intervention. Neuroblastoma is consistently regarded as one of the most common cancers to undergo spontaneous regression, in addition to carcinoma of the kidney, malignant melanoma, choriocarcinoma, and lymphoid malignancies.^49–52 However, the prevalence of neuroblastoma regression was unknown until recently.

Beckwith and Perrin studied the adrenal glands of infants less than 3 months old who were not known to have neuroblastoma and had died for various reasons. They found microscopic foci of neuroblastic cells in the adrenal glands of 1 out of 40 infants studied, and they suggested the term “in situ neuroblastoma” to describe this phenomenon.⁵³ The researchers proposed that these neuroblastic nodules might eventually evolve into clinically detectable neuroblastoma. Based on this assumption and the normal prevalence of clinically detected neuroblastoma of about 1 in 8,000 live births, the researchers inferred that the prevalence of spontaneous regression of neuroblastoma was about 200-fold higher than clinically detected disease. However, subsequently others have studied the adrenal glands of foetuses that were spontaneously aborted, and found that similar neuroblastic foci were present in all foetuses, with a peak time of appearance between 16–20 weeks of gestation, and after that these foci gradually disappeared.^54,55 Thus, it is likely that neuroblastic rests seen in the first study were not insipient nodules of malignant neuroblastoma in situ, but rather the residual elements of normal sympathoadrenal development. Nevertheless, these normal neuroblastic nodules might contain the cells from which adrenal neuroblastomas arise.

Neuroblastoma stage 4S

The phenomenon of spontaneous regression of neuroblastoma was already known, but it was further highlighted by Evans and D’Angio who identified a specific pattern of metastatic spread called stage IVS.^56,57 Infants with stage IVS generally had small primary tumours with dissemination limited to the liver and skin, with minimal bone marrow involvement. These patients were noted to have a very good prognosis, and some did well even in the absence of any tumour-specific therapy.^56,57 This is in contrast to patients over 12–18 months of age with a different pattern of metastatic disease that included bone lesions and more-extensive marrow involvement. Patients with the latter pattern of metastatic disease generally had a poor prognosis.^58,59 Later, the definition of stage IVS was refined as stage 4S by the International Neuroblastoma Staging System (INSS).^60,61 INSS stage 4S was restricted to infants aged less than 12 months at diagnosis, with small primary tumours (INSS stage 1 or 2), and with less than 10% marrow involvement. The most recent staging system iteration is called the International Neuroblastoma Risk Group Staging System (INRGSS), and this stage is referred to as stage MS (M for metastatic, S for special); ⁶² we will refer to this stage as 4S in this Review. However, it is clear that spontaneous regression is not restricted to this stage or subset of patients, as regression can be seen in infants with any stage of disease if they have biologically favourable tumours.^48,63–65

Genomics and biology of 4S tumours

A limited number of studies have been performed to identify patterns of gene expression that are characteristic of stage 4S neuroblastomas. Benard et al.⁴⁶ studied 29 cases of metastatic neuroblastoma (12 of which were stage 4S) that lacked MYCN amplification. The investigators developed a genetic signature of 45 genes that was significantly associated with stage 4S versus stage 4 tumours, and this was validated in an independent set of 22 tumours. Lavarino et al.⁶⁶ conducted a similar study on 35 infants with metastatic neuroblastoma (25 had stage 4 disease and 10 had stage 4S neuroblastoma). The tumours from patients with stage 4 disease were characterized by SCAs, whereas 90% of stage 4S tumours were near-triploid with whole chromosome gains.⁶⁶ The investigators found a differential expression of certain genes (such as CHD5, GNB1 and RERE) in stage 4 and stage 4S neuroblastoma, with higher expression of genes mapping to the short arm of chromosome 1 in stage 4S tumours, and to chromosome 11 for stage 4 tumours. A smaller proteomic study was performed on eight tumours from infants with stage 4 and 4S, which identified another set of differentially expressed proteins across the two stages.⁶⁷ However, there was essentially no overlap of genes (or proteins) differentially expressed by regressing 4S versus non-regressing infant tumours among these studies, suggesting that further investigation is needed.

Metastatic and multifocal neuroblastic tumours

The unique pattern of metastasis characteristic of stage 4S disease might represent multifocal hyperplasia of neural crest precursors.^68,69 Indeed, the tissues involved in the 4S dissemination pattern are reminiscent of the tissue migration pattern of neural crest cells observed during development (Box 1). The concept of multifocal hyperplasia is based on the hypothesis that all cells have sustained a single initiating lesion, whereas true malignancies have sustained two or more critical mutations.⁶⁸ In this regard, it is interesting that patients with constitutional activating mutations of ALK and hereditary neuroblastoma do not present a 4S pattern with increased frequency.^20,24–26 To date, no biopsy or genotype analysis has been performed on multiple discrete nodules from patients with 4S tumours to evaluate if those patients have identical or non-identical patterns of genomic changes. However, genomic analysis of primary 4S tumours reveal clonal changes, consistent with a malignant process.^47,66,70,71 Genomic analyses of multiple lesions in 4S patients would likely address this issue definitively; however, considering the available data, we favour the hypothesis that stage 4S tumour are metastatic dissemination of genetically identical cells, developing as biologically favourable disease in most cases.

Box 1. Genes involved in neural crest development.

Insights into the mechanisms of neuroblastoma regression could potentially be gained by understanding the developmental biology of the neural crest, from which neuroblastomas are derived. During embryonic development, cells at the edge of the neural plate undergo epithelial-to-mesenchymal transition and migrate away from the neural tube.¹⁶⁴ These cells follow different pathways, ultimately giving rise to a variety of tissues, including the enteric nervous system, melanocytes, dorsal root and sympathetic ganglia, and the chromaffin cells of the adrenal medulla. Numerous transcription factors and signalling molecules contribute to neural crest formation and to the differentiation of pluripotent neural crest cells.¹⁶⁵ Furthermore, epigenetic control might also be important in determining cell fate.¹⁶⁴ Much remains to be learned regarding genes normally involved in neural crest development and their role in neural crest-derived malignancies, such as neuroblastoma. However, expression profiling of a relatively small number of foetal adrenal neuroblasts, normal foetal adrenal cortical cells, and neuroblastoma tumours has suggested that there is differential expression of genes associated with earlier (ASCL1) versus later (DLK1) time points in neural crest development in tumours classified as favourable versus unfavourable.¹⁶⁶ Furthermore, MASH1, PHOX2B, Hand2 BMP4 and other genes are known to have critical roles in chromaffin, dorsal root and sympathetic neural development.^167,168 More recently, studies in zebrafish have confirmed that PHOX2B is required for normal development of neural crest-derived cells that comprise the sympathetic ganglia,¹³⁶ and that specific PHOX2B aberrations can have varying effects on differentiation.¹⁶⁹ Additional studies could potentially clarify if PHOX2B and other genes involved in neural crest development have a role in neuroblastoma regression as well.

Lessons from neuroblastoma screening

The exact prevalence of spontaneous regression has been difficult to define precisely, but neuroblastoma mass-screening studies undertaken in Japan, Canada and Europe have provided useful information. The outcome of infants with neuroblastoma is substantially better than that of older children with this disease. Furthermore, almost all neuroblastomas produce catecholamines, and their metabolites—homovanillic acid (HVA) and vanillylmandelic acid (VMA)—can be readily detected in urine.⁷² Therefore, mass screening of infants for neuroblastoma by measuring urine VMA and HVA was initiated in Japan,^11,12 with the goal of detecting tumours earlier and hopefully improving patient outcome. Initial results were promising, so similar efforts were initiated in North America and in Europe.^13,14 Analysis of the data generated from these initiatives indicates that mass screening resulted in a substantial increase in the prevalence of neuroblastoma in the screened population (~1:2,000 live births, versus 1:7,000 live births in unscreened populations), but overall mortality was unchanged. Importantly there was no significant decrease in the prevalence or mortality of neuroblastoma in patients older than 1 year.^73–76 Thus, the goal to reduce neuroblastoma mortality by mass screening in infancy was not achieved, and screening efforts have essentially stopped worldwide.

Despite the fact the main objective of mass-screening was not reached, studies of this approach have provided valuable insights into the pathogenesis and clinical behaviour of biologically favourable tumours. The increased prevalence of neuroblastoma observed in the screened populations indicates that spontaneous regression of neuroblastoma (without clinical detection) occurs at least as frequently as neuroblastoma that is detected clinically. In addition, analyses performed on screened tumours showed that virtually all of them, regardless of their stage, were biologically favourable with respect to MYCN status and tumour cell ploidy,^77–79 which is in contrast to the unfavourable biological features generally found in clinically detected tumours from older children. Importantly, these studies also indicated that biologically favourable tumours rarely evolve into biologically unfavourable tumours.

Although not intended as a neuroblastoma screening tool, there have been several reports of incidental prenatal detection of neuroblastoma by maternal ultrasound.^80–82 These cases are similar, both clinically and biologically, to those identified by screening, and the vast majority of infants do well with little or no therapy. Infants under 1 year of age are at risk for both anaesthetic and surgical complications associated with resection of adrenal tumours,⁸³ and so a strategy of expectant observation has been evaluated. Consistent with the findings from previous studies,^84–86 a recent cooperative group trial demonstrated that more than 80% of such patients who had an adrenal mass detected in the perinatal period were spared surgical intervention, and the overall survival in this group was 100%.⁸⁷ Thus, careful observation is now considered standard care in young infants (up to 6 months of age) who have small tumours found incidentally.

Mechanisms of spontaneous regression

Neurotrophin receptors and regression

The Trk neurotrophin receptors—TrkA/NTRK1, TrkB/ NTRK2, and TrkC/NTRK3—have critical roles in the development and maintenance of the central and peripheral nervous systems. The cognate ligands for these receptors are nerve growth factor (NGF), brain-derived neurotropic factor (BDNF) and neurotrophin-3 (NT3) growth factor, respectively. These receptors also have important roles in neuroblastoma pathogenesis.^88–90 High TrkA expression levels are associated with favourable clinical and biological features, such as younger age, lower stage, and absence of MYCN amplification, and these patients have an excellent outcome.^91–94 By contrast, TrkB is coexpressed at high levels with its ligand, BDNF, in clinically and biologically unfavourable tumours, especially those with MYCN amplification.⁹⁵ Activation of the TrkB–BDNF autocrine pathway can lead to invasion, metastasis, angiogenesis and drug resistance.^95–99 The TrkA and TrkC receptors are also known as dependence receptors, as the absence of ligand activation will generate apoptotic signals.^100,101 Coexpression of the P75/NGFR receptor can increase the sensitivity and specificity of all three Trk receptors for their cognate ligands;^102,103 however, overexpression and activation of P75/NGFR in the absence of Trk expression can lead to apoptosis.^101,104

Tumours from patients with low-stage and 4S disease generally express high levels of TrkA.^91–94 When cells derived from these tumours were put in primary culture in the presence of NGF, they underwent neuronal differentiation, and survived for months. By contrast, the same cells died via apoptosis within a week once deprived of NGF.^93,105 Thus, these in vitro culture conditions seem to recapitulate the behaviour of TrkA-expressing neuroblastomas in patients whose tumours undergo neuronal differentiation or spontaneous regression (apoptosis), depending on the presence or absence of NGF in their microenvironment (Figure 2; Box 2).

Figure 2.

Mechanisms of spontaneous regression. Shown are the major pathways that have been proposed to explain the phenomenon of spontaneous regression. These include: neurotrophin deprivation (TrkA without NGF) and activation of developmentally programmed apoptosis (1); immune-mediated cell killing by anti-neuroblastoma antibodies (and antibody-dependent cellular toxicity) or by NK cells (2); telomere shortening and apoptosis triggered by low/absent levels of telomerase (3); and epigenetic changes in gene expression controlled by DNA methylation, histone modification, or alterations in chromatin remodelling (4). Abbreviations: NGF, nerve growth factor; NK, natural killer cells; TrkA, tropomyosin receptor kinase A.

Box 2. Mechanisms of neuroblastoma differentiation.

Neuroblastic tumours consist of a histological spectrum that ranges from immature neuroblastoma and differentiating ganglioneuroblastoma, to ganglioneuroma. The extent of differentiation in the tumour has prognostic significance.^{19,161,170,171} Ganglioneuromas are comprised of individual ganglion cells or clusters surrounded by Schwannian stroma, but there is controversy concerning the origin of the Schwann cells in these tumours. Ambros et al.¹⁷² showed that the Schwann cells in ganglioneuromas had a normal DNA content, whereas the ganglionic cells were clonally aneuploid, suggesting the Schwann cells are normal, infiltrating host cells. Mora et al.¹⁴² showed that the stromal cells in treated neuroblastomas shared the same genomic abnormalities as immature neuronal components, suggesting they were part of the malignant process. Ambros and colleagues¹⁷² suggested that the Schwann cells were recruited into a less mature tumour by a Schwann-trophic factor, and the infiltrating Schwann cells in turn proliferated and produced NGF as well as other neurotrophins that caused an immature neuroblastoma to differentiate into a benign ganglioneuroma.¹⁷³ Indeed, ganglioneuromas usually have abundant expression of TrkA, so they would be particularly prone to differentiate in the presence of NGF. Even immature neuroblastomas can undergo neuronal maturation into ganglionic cells in vitro in the presence of NGF.^93,105 The identity of the Schwann-trophic factor (or factors) is unknown, and it is unclear why some tumours produce such factors, whereas others do not. However, ganglioneuromas are rare in infants, so this might represent a feature of favourable, TrkA-expressing tumours in older patients. Therapeutic approaches aimed at inducing differentiation in neuroblastomas have been suggested. Treatment with agents such as NGF might be effective at inducing differentiation in TrkA-expressing tumours. However, an alternative approach might be to induce expression of Schwann-trophic factors, resulting in tumour infiltration and differentiation via the neurotrophic factors produced by Schwann cells.

Abbreviations: NGF, nerve growth factor; TrkA, tropomyosin receptor kinase A.

Migrating neural crest precursors and favourable neuroblastomas that express TrkA on their cell surface survive (at least initially), despite a lack of available NGF, and the reason for this paradox is uncertain. One possible explanation comes from the identification of TrkAIII, a TrkA isoform that is expressed in normal sympathoadrenal progenitors as well as in some neuroblastomas. ^106,107 This isoform results from the alternative splicing of exons 6, 7 and 9, which maintains the reading frame but abolishes the ligand-binding site, leading to constitutive activation of the kinase. Thus, the conversion from a TrkA-expressing, NGF-dependent neuroblastoma to a NGF-independent one could be the consequence of a developmentally programmed isoform switch from TrkAIII to TrkAI. Alternatively, NGF-independent neuroblastomas could depend on another receptor or pathway for survival (such as TrkC or RET),^108–110 and then switch dependence to TrkA, only to undergo apoptosis and regress in the absence of ligand. However, the available data regarding the possible mechanisms of neuroblastoma regression are most consistent with a TrkA dependence mechanism.

Immunological mechanisms and regression

Spontaneous regression of neuroblastomas and other types of cancer is sometimes associated with an acute infection. Regression of experimental tumour metastases can also be induced by immunomodulatory cytokines.^111,112 Furthermore, tumour-infiltrating lymphocytes are observed occasionally in neuroblastomas, and there is evidence for the presence of both tumour-targeted T-cells and antineural antibodies in patients with neuroblastoma.^113–115 Thus, a plausible theory is that spontaneous regression can be a consequence of a host-mediated immune response (Figure 2). Interestingly, the paraneoplastic opsomyoclonus syndrome (OMS) is associated with the presence of antineural antibodies and a favourable outcome in patients with neuroblastoma.^{113,116–118} About 50% of patients with OMS have neuroblastoma, which suggests that the other 50% either had a neuroblastoma that regressed, or they have a de novo autoimmune disease. Unfortunately, the OMS can persist or recur long after removal and eradication of the tumour, so these patients might benefit from additional immunosuppressive therapy.¹¹⁹

Neuroblastoma cells from patients with high-risk disease have been shown to evade immune destruction by downregulating human leucocyte antigen (HLA) class I molecules.¹²⁰ However, most tumours from patients with stage 4S neuroblastoma express normal levels of HLA class I antigens.¹²¹ Upregulation of the expression of class I antigens in neuroblastoma cells can be induced in vitro by incubating cells with interferon-γ (IFN-γ).¹²⁰ This raises the possibility that upregulation of HLA class I in vivo might represent a strategy to augment immune surveillance and ultimately promote tumour regression. However, although HLA class I upregulation following in vitro exposure to IFN-γ has been shown to enhance the recognition of neuroblastoma cells by cytotoxic T cells, it could also reduce their susceptibility to killing by natural killer (NK) cells.¹²⁰ Alteration in the expression of components of the antigen-processing machinery (including HLA class I heavy chain) have been demonstrated in primary tumours from patients with neuroblastoma, and may contribute to decreased killing of neuroblastoma cells.¹²⁰ Further study of the role of these molecules in tumour regression is needed, as expression of these immune components has been studied in only a limited number of low stage and 4S neuroblastomas.

A study designed to assess the tumour-associated macrophage (TAM) infiltration in tumours from patients with various stages of MYCN non-amplified disease showed that INSS stage 4 tumours contained significantly increased numbers of CD163+ cells; however, this study reported that the number of TAM in INSS 4S tumours is similar to locoregional neuroblastomas.¹²⁶ Metastatic tumours from young patients (<18 months old) had significantly higher expression levels of genes representing TAMs (CD33, CD16, IL6R, IL10, FCGR3) than did tumours from patients ≥18 months of age. This suggests that the inflammatory response and the tumour microenvironment might have important effects on the natural history and outcome of neuroblastoma in specific patient groups.¹²⁶ This study was conducted on a small cohort, which likely limited the investigators’ ability to evaluate expression of TAM-associated genes in patients with regressing tumours, and highlights the need for further study on the role of the immune system and the tumour microenvironment in the context of disease regression.

Telomerase, telomeres and regression

Telomeres are specialized structures at the ends of chromosomes that are involved in the replication and stability of the chromosome itself. They have an important role in guaranteeing genomic stability and are in a state of dynamic equilibrium. The regulation of the telomere length is controlled in part by the enzyme telomerase. Of note, telomerase expression is frequently high in cancer and immortalized cells, but low in most normal and senescent cells.¹²⁷ Hiyama and colleagues¹²⁸ studied the regulation of telomere length and the activity of the telomerase in 100 samples of neuroblastomas. Most of the tumours that exhibited a high level of telomerase activity were associated with a poor prognosis, and all tumours with MYCN amplification had high telomerase activity. Interestingly, most of the tumour samples from 4S neuroblastoma had low telomerase activity or short telomeres, a pattern that is usually associated with senescent cells (Figure 2).¹²⁸ Furthermore, Samy et al.¹²⁹ transfected a neuroblastoma cell line (IGR-N-91) with a dominant negative form of human telomerase, h-Tert. The h-Tert-transfected neuroblastoma cells formed tumours that showed more apoptosis and reduced tumorigenicity in a mouse xenograft model compared to untransfected neuroblastoma cells.

These data suggest that loss of telomerase activity is a plausible mechanism to explain spontaneous regression of neuroblastoma, and possibly of other tumours. However, low telomerase activity is associated with biologically favourable tumours, which also have hyperdiploidy, high expression of TrkA and lack high-risk genomic features, such as MYCN amplification.¹³⁰ Thus, the role of telomerase is not clear; the association between stage 4S and regression might be related to the low telomerase activity or might depend on other favourable features associated with stage 4S, such as younger age, hyperdiploidy and lack of MYCN amplification. Nevertheless, high levels of telomerase activity is generally associated with a more-aggressive tumour behaviour and a poor prognosis in patients with neuroblastoma.^131–133

Epigenetic regulation and other mechanisms

Changes in gene expression related to alterations in promoter methylation, histone modification or chromatin remodelling might also impact differentiation in neuroblastoma cells. Epigenetic changes affecting expression of genes relevant to neuroblastoma development were initially reported more than a decade ago,^134,135 and several studies have suggested that alterations in gene methylation and histone modification are related to patient outcome (Figure 2).^136–139 The correlation between epigenetic changes and neuroblastoma behaviour has been increasingly studied, particularly because next-generation sequencing analyses of neuroblastoma have reported a very limited number of previously unrecognized recurrent somatic mutations.^41–44 Furthermore, data from other childhood tumour types, particularly Wilms tumour and medulloblastoma, suggest that epigenetic changes might help to explain poorly understood aspects of disease presentation and clinical behaviour.^140–142

The development of genome-wide methylation detection methods has facilitated study of epigenetics in numerous tumour types, and in neuroblastomas specifically.¹³⁹ Preliminary data showed that there are global differences in the methylomes of 22 neuroblastoma stage 4S tumours compared to the methylomes of low-risk and high-risk tumours and to the methylome of normal brain tissues.⁴⁷ In the 4S samples, reduced promoter methylation was observed in 97% of the genes in which differential methylation was detected. Specifically, differentially methylated promoters were enriched for genes known to have binding sites for transcription factors involved in cell differentiation (such as LEF1, TCF3, ETS2, and PITX2).⁴⁷ Other investigators have also reported different patterns of methylation in tumours from patients with 4S versus other disease stages,¹³⁹ but additional studies are needed to confirm and extend these initial findings. Studies of agents that affect DNA methylation status, histone modification, or chromatin modifiers during differentiation and regression are ongoing,^143–145 but currently there are no clinical trials of epigenetic modifiers in stage 4S neuroblastomas.

Therapeutic implications of regression

Targeting the TrkA pathway

One of the most promising approaches for the induction of apoptosis and tumour regression in neuroblastomas is targeting the TrkA neurotrophin receptor pathway. This is based on the observation that TrkA-expressing tumour cells placed in culture will survive and even differentiate in the presence of NGF, but undergo apoptosis in its absence.¹⁰⁵ Thus, depriving cells of NGF or inhibiting TrkA signalling might be an effective approach to induce regression. Lestaurtinib (CEP-701) is a small molecule that targets Trk neurotrophin receptors (TrkA, TrkB and TrkC), and it has shown preclinical activity against TrkB-expressing neuroblastoma xenografts.^146–149 Furthermore, lestaurtinib, when given at biologically effective doses, has shown significant clinical activity in a phase I trial in children with recurrent and/or refractory neuroblastoma.¹⁵⁰ Lestaurtinib is not being supported for future clinical trials, but these studies provide proof-of-principle that Trk-selective inhibitors could be effective in the treatment of neuroblastomas driven by Trk receptors. Indeed, several second-generation Trk inhibitors are currently in phase I clinical trials or in preclinical development.^151–154 These agents are potent inhibitors of all three Trk family neurotrophin receptors, so the same agent could be used to target TrkA in favourable 4S tumours, and TrkB in unfavourable tumours.

In general, new agents are tested in patients with recurrent and/or refractory disease; therefore, new Trk inhibitors will not be tested initially in infants with de novo stage 4S disease or other patients with locoregional tumours and biologically favourable features. However, there is significant mortality due to respiratory compromise among very young infants with stage 4S neuroblastoma and hepatomegaly.^155,156 If second-generation Trk inhibitors prove to be safe and effective against TrkB-expressing, high-risk disease, it would be reasonable to consider administration of these agents in patients with TrkA-expressing 4S neuroblastoma and massive liver involvement in lieu of chemotherapy or radiation therapy. Theoretically, a Trk inhibitor could initiate the process of apoptosis and regression in these tumours. It would be highly desirable to have an agent, such as a Trk inhibitor, that can induce regression directly, rather than waiting for it to occur spontaneously. Trk inhibitor therapy might also be used for infants with large abdominal tumours or ‘dumbbell tumours’ with intraspinal extension, to spare such patients the toxicity and long-term side effects of laminectomy, chemotherapy or spinal radiation.

Immunological approaches

Immunotherapy using a chimeric antibody (ch14.18) directed against the disialoganglioside GD2 has been incorporated into frontline treatment of patients with high-risk neuroblastoma,¹⁵⁷ and very preliminary studies have been performed using adoptive immunotherapy in patients with relapsed and refractory disease.¹⁵⁸ Additional studies of the immunology of differentiation and regression are expected to influence the evolution of current immunotherapeutic approaches and could potentially result in new strategies to accelerate regression in young infants with acute life-threatening, but biologically favourable disease. Immune modulation to induce regression could potentially be advantageous for patients with 4S or locoregional disease, and a trend in the field has been to reduce conventional cytotoxic therapy and avoid aggressive surgery in patients with favourable prognoses.^84,87,159 The toxicities associated with currently available immunocytokine therapies are considerable, and little is known regarding late effects of these therapies, especially in very young infants. Less toxic immune modulation therapy could be considered, including therapy designed to enhance immune surveillance by increasing HLA I expression on neuroblastoma cells. However, as noted above, strategies to enhance one component of the immune system may diminish the antitumour effects of another key component. There is an increasing need for a better understanding of the complex interactions between neuroblastoma cells and the immune system, and of the multiple implications of immune modulation in very young children.

Other approaches

At the present time, there are no therapeutic approaches available to influence telomere length in neuroblastomas. However, in addition to targeted inhibition of the TrkA pathway or immunological therapy, there are other approaches that might be considered for these patients, For example, the retinoids are a class of compounds that have been shown to induce cellular differentiation and decrease proliferation of neuroblastoma cells in vitro, presumably mediated by the induction of expression of neural differentiation genes.¹⁶⁰ Indeed, the retinoid isotretinoin (13-cisretinoic acid) has been incorporated into frontline therapy for children with high-risk neuroblastoma in an effort to induce differentiation in vivo in states of minimal residual disease following intensive, multimodality therapy.¹⁶¹ The precise mechanisms by which isotretinoin induces differentiation are unclear, but it seems that retinoids are associated with increased expression of Trk receptors.¹⁶⁰

Vorinostat, a histone deacetylase inhibitor, has been administered in combination with isotretinoin in children with relapsed or refractory disease in an effort to further induce neuroblastoma cell differentiation. A patient with neuroblastoma who had evaluable disease experienced a complete response to therapy on one phase I trial,¹⁶² although no objective responses were seen in a second study.¹⁶³ Further studies of epigenetic changes in neuroblastoma cells that undergo differentiation versus those that do not differentiate could lead to the development of more specific and effective agents that can alter gene expression in neuroblastomas and consequently induce tumour regression.

Conclusions

Neuroblastomas show a remarkable capacity to undergo spontaneous regression. The prevalence of this phenomenon is hard to determine precisely, but the experience from mass-screening programmes suggests that there are at least as many children who have tumours undergoing spontaneous regression without detection as there are patients with neuroblastoma detected clinically. Further exploration of this issue and a greater understanding of the normal mechanism(s) of spontaneous regression might allow the identification of tumours that have the capacity to undergo spontaneous regression and to induce regression in susceptible tumours using pharmacological, biological or immunological approaches. However, to this end, we will need to study samples from a substantial number of regressing tumours, or perhaps establish an animal model that suitably mimics the process of spontaneous regression. At the present time, the most promising therapeutic approach would be aimed at inhibiting the TrkA receptor pathway. However, most Trk inhibitors are potent inhibitors of TrkA, TrkB and TrkC. Before these agents are used to treat infants with stage 4S disease, clinical trials of second- generation Trk inhibitors would need to demonstrate safety and efficacy against TrkB-expressing recurrent and/or refractory neuroblastomas.

Review criteria.

We searched for original articles focusing on neuroblastoma regression in MEDLINE and PubMed published from 1971 to June 2014. In terms of potential mechanisms, we focused on articles that had been published in the past 15 years. The search terms used were “neuroblastoma” and one or more of the following: “regression”, “spontaneous regression”, “spontaneous differentiation”, “telomerase”, “regression and mechanisms”, “regression and telomerase”, “regression and TRK”, “epigenetics”, “immune function”, “NK cells”, “regression and immune”. We also searched “genes” and “neural crest development” and “neuroblastoma and mass screening”. All papers identified were English-language full-text papers. We also searched the reference lists of identified articles for further papers.

Key points.

• Neuroblastomas in younger children have a higher propensity to undergo spontaneous regression than any other human malignancy, but only genetic subtype 1 tumours with numerical chromosome aberrations are prone to spontaneous regression

• Although spontaneous regression of neuroblastoma is strongly associated with stage 4S disease, mass-screening studies suggest that genetic subtype 1 tumours of any stage in infants <18 months can undergo spontaneous regression

• The TrkA/NGF pathway might have a major role in causing spontaneous regression, but alternative mechanisms might involve cellular or humoral immune responses, telomere shortening, or epigenetic modifications

• Inhibition of the TrkA pathway represents the most-promising approach to initiate regression in infants with favourable tumours; other approaches involving immune modulation or epigenetic regulation are being investigated