Involvement of midkine in neuroblastoma tumourigenesis

Authors


Abstract

Midkine is highly expressed in various cancers, including neuroblastoma, one of the most malignant paediatric solid tumours known. Also, it has been shown to be useful as a tumour marker, a prognosis factor and a target of molecular therapy. Several molecular tools (e.g. siRNA, antibodies and RNA aptamer) have been used to establish a midkine-targeted therapy. The involvement of midkine in tumourigenesis has been demonstrated in vivo in a mouse neuroblastoma model, where targeting it with an RNA aptamer was shown to be an effective treatment for xenografted tumours. Chemoresistance is one of the notable phenotypes regulated by midkine in various cancer cell types. In pancreatic tumours and glioma cells, midkine is expressed in chemoresistant cells and is involved in the survival of these cells in the presence of anticancer drugs. In contrast to these tumours, midkine was found to be expressed in every neuroblastoma cell line tested and the knockdown of midkine alone was sufficient to suppress their growth. These results indicate that neuroblastoma cells are highly dependent on midkine and that a midkine-targeted therapy could exert a significant effect in these cells. However, to achieve a midkine-targeted therapy for high-risk neuroblastoma patients, the further refinement of the RNA aptamer or antibody as tools and the elucidation of midkine signalling are immediate issues that need to be resolved. Regarding the latter, although it has been shown that Notch2 functions as a receptor in neuroblastoma cells, it is likely that other receptors (e.g. anaplastic lymphoma kinase) are also involved in midkine signalling.

Linked Articles

This article is part of a themed section on Midkine. To view the other articles in this section visit http://dx.doi.org/10.1111/bph.2014.171.issue-4

Abbreviations
ALK

anaplastic lymphoma kinase

EMT

epithelial-mesenchymal transition

HES1

hairy and enhancer of split 1

LRP1

LDL receptor-related protein 1

PDAC

pancreatic ductal adenocarcinoma

SG

sympathetic ganglion

SMG

superior mesenteric ganglion

Tg

transgenic

TH

tyrosine hydroxylase

Midkine in various cancers

Midkine is a heparin-binding growth factor identified over 20 years ago (Kadomatsu et al., 1988; Tomomura et al., 1990). Midkine was discovered through the screening of differentially expressed genes during the differentiation of embryonal carcinoma cells (Kadomatsu et al., 1988). Midkine protein, whose molecular weight is 13 kDa, is rich in basic amino acids and cysteines, and shares around 50% sequence identity with another heparin-binding growth factor, pleiotrophin (Kadomatsu et al., 1988; Tomomura et al., 1990). These two compounds comprise the entire midkine family of growth factors. The midkine protein is composed of two domains, the N- and C-terminal half domains respectively. Each domain contains three anti-parallel β sheets that are linked by a flexible linker region. Interestingly, only the C-terminal-half domain of midkine exhibits biological activities, for example neurite outgrowth, fibrinolysis and nerve cell migration.

So far, it has been shown that midkine is involved in a huge variety of biological phenomena, including development, inflammation, tissue protection and blood pressure (Kadomatsu and Muramatsu, 2004; Kadomatsu et al., 2013). For example, ectopically expressed midkine can transform NIH3T3 cells in vitro (Kadomatsu et al., 1997). One of the most notable characteristics of midkine is its significant expression in a variety of cancers. It has been shown to be highly expressed in Wilms' tumour (Tsutsui et al., 1993), gastrointestinal cancer (Aridome et al., 1995), astrocytoma (Mishima et al., 1997), colorectal cancer (Ye et al., 1999), prostate cancer (Konishi et al., 1999) and neuroblastoma (Nakagawara et al., 1995). In contrast, in healthy animals its expression is mostly restricted to midgestation during embryogenesis (Kadomatsu et al., 1990). These findings strongly suggest that midkine would be useful not only as a tumour marker but also as a molecular target for cancer therapy. And indeed, serum midkine levels have been shown to be a reliable tumour marker, as well as a prognostic factor, in several cancers, including neuroblastoma (Ikematsu et al., 2000; 2003; 2008; Obata et al., 2005; Jia et al., 2007; Ota et al., 2008; Ibusuki et al., 2009). In cases of neuroblastoma, the plasma midkine level is significantly correlated with known prognostic factors, such as MYCN amplification, low TRKA expression, diploidy, stage 3 and 4 disease, and over 18 months old at diagnosis. In addition, there is a striking correlation between a high plasma midkine level itself and poor prognosis (Ikematsu et al., 2003; 2008).

With regard to the use of midkine as a molecular target in cancer therapy, several molecular tools have been established to target midkine, and have been shown to exert therapeutic effects on xenografted tumours derived from several types of cancer (Table 1). These tools can be mainly classified into three groups: (i) RNAi-based nucleotides (Takei et al., 2001; 2002; 2005; 2006; Dai et al., 2007a,b; 2009; Lorente et al., 2011); (ii) small peptides or antibodies (Chen et al., 2007; Sueyoshi et al., 2012); and (iii) RNA aptamers (Kishida et al., 2013). They are expounded in the next section.

Table 1. Cancer therapies targeting midkine
ToolsExperimentsReferences
OligodeoxynucleotideIntratumour injection to CMT-93 (mouse rectal carcinoma) xenograft

Takei et al., 2001

Takei et al., 2002

Morpholino oligomerIntratumour injection to PC-3 (prostate carcinoma) and SW620 (colon carcinoma) xenograftsTakei et al., 2005
siRNAIntratumour injection to PC-3 xenograftTakei et al., 2006
OligonucleotideI.v. injection to in situ hepatocellular carcinoma modelDai et al., 2007a
OligonucleotideI.v. injection to in situ hepatocellular carcinoma modelDai et al., 2007b
MK-TRAP (midkine-binding peptide derived from LRP1)MK-TRAP-transfected CMT-93 xenograftChen et al., 2007
Polyclonal antibodyAnchorage-independent colony formation of G401 (Wilms' tumour) and CMT-93Chen et al., 2007
Oligonucleotide-loaded nanoparticleI.v. injection to in situ hepatocellular carcinoma modelDai et al., 2009
siRNAIntratumour injection to T98 (glioma) xenograftLorente et al., 2011
Monoclonal antibodyI.p. injection to 143B (osteosarcoma) xenograftSueyoshi et al., 2012
RNA aptamerIntratumour injection to TNB1 and YT-nu (neuroblastomas)Kishida et al., 2013

Involvement of midkine in neuroblastoma

Although it has been reported that midkine is highly expressed in a variety of cancers (Tsutsui et al., 1993; Aridome et al., 1995; Nakagawara et al., 1995; Mishima et al., 1997; Konishi et al., 1999; Ye et al., 1999) and that its suppression results in growth inhibition of tumour cells (Takei et al., 2001; 2002; 2005; 2006; Chen et al., 2007; Dai et al., 2007a,b; 2009; Lorente et al., 2011; Sueyoshi et al., 2012; Kishida et al., 2013), no direct or conclusive evidence shows that midkine is involved in tumourigenesis. In order to address this point, in vivo data utilizing animal models is needed. Recently, the involvement of midkine in neuroblastoma, one of the most malignant paediatric solid tumours known, was revealed for the first time in a study using a mouse model (Kishida et al., 2013).

Neuroblastoma develops from a neural crest-derived sympathetic neuronal lineage, and accounts for around 15% of all paediatric cancer deaths (Nakagawara and Ohira, 2004). Its prognosis still remains poor despite an enormous amount of basic and clinical research. However, neuroblastoma is also famous for the unique characteristics of a subclass of tumours classified as stage 4S (‘S’ stands for ‘special’). The tumours in patients with stage 4S disease (with metastases in the liver, skin or bone marrow) subsequently disappear through apoptosis and/or differentiation without any treatment. In other words, stage 4S patients can be free from malignancy provided they can withstand the life-threatening period when the tumours are large. This phenomenon is called spontaneous regression, and its molecular mechanisms are almost entirely unknown. It is very important for us to investigate and understand the mechanism of spontaneous regression, since the phenomenon could provide clues to the development of novel therapies for malignant neuroblastomas other than stage 4S. Midkine mRNA is highly expressed in neuroblastoma tissues with poor prognosis, whereas pleiotrophin, another family member of midkine, is expressed in tumours with good prognosis (Nakagawara et al., 1995). Although these two proteins seem to have redundant functions in some fields, this does not appear to be the case with neuroblastoma. As mentioned above, increased serum midkine levels appear to be associated with poor prognostic factors in human neuroblastoma (Ikematsu et al., 2000; 2003; 2008). The oncogenic transcription factor MYCN, which is a member of the same family as c-Myc, is considered to be the most important gene indicative of a predisposition for the development of neuroblastoma, and its amplification is one of the most important prognosis factors so far. The MYCN gene initially plays a role in the normal development of sympathetic neurons (Nakagawara and Ohira, 2004), but the ectopic expression of MYCN at a particular stage of development is thought to be a trigger for oncogenesis, and as a result, the neuroblasts in which human MYCN is ectopically expressed remain in an undifferentiated state and proliferate ad infinitum. MYCN transgenic (Tg) mice in which the human MYCN gene was introduced under the control of a rat tyrosine hydroxylase (TH) promoter have been developed as an animal model for neuroblastoma, (Weiss et al., 1997). TH is a sympathetic neuron-specific enzyme, and its promoter is turned on in neural crest-derived migrating cells whose lineages are committed to sympathetic neurons during early development. MYCN Tg mice spontaneously develop tumours from the superior mesenteric ganglion (SMG), one of the sympathetic ganglia. The SMG lies on the superior mesenteric artery between the left and right kidneys, and its size is less than 1 mm in diameter. Although the most frequent origin of human neuroblastoma is the adrenal grand, the dominant origin in MYCN Tg mice is the SMG. These tumours are pathologically similar to human neuroblastoma with respect to both histology (Weiss et al., 1997) and pattern of chromosomal aberration (Weiss et al., 2000) and they have been utilized for a variety of basic investigations of neuroblastoma. At the age of 2 weeks, the SMG in wild-type mice consists of fully differentiated sympathetic ganglion cells, whereas in all MYCN Tg mice undifferentiated neuroblasts are accumulated in the SMG (Hansford et al., 2004; Asano et al., 2010; Huang et al., 2011). These cells are collectively referred to as a hyperplasia. The hyperplasia might be a kind of precancerous stage, because most hyperplasia develop into advanced tumours. In MYCN Tg mice, midkine was highly expressed in both those precancerous SMG and later terminal tumour tissues (Kishida et al., 2013). Next, in order to directly determine the involvement of midkine in the tumourigenesis of MYCN Tg mice, Mdk (the mouse gene name of midkine)-knockout mice were crossed with MYCN Tg mice. As a result, it was found that the genetic ablation of Mdk resulted in suppressed tumourigenesis (lower tumour incidence and delayed growth) of MYCN Tg mice (Kishida et al., 2013). This is the first study to investigate the involvement of midkine in tumourigenesis at the in vivo level, and the results demonstrate that midkine has certain properties that promote the tumourigenesis of neuroblastoma.

It should be noted that the mechanism of action of midkine has not been fully elucidated. In terms of the initiation of midkine signalling, several receptor candidates, including anaplastic lymphoma kinase (ALK; Stoica et al., 2002), Notch2 (Huang et al., 2008; Güngör et al., 2011; Kishida et al., 2013), LDL receptor-related protein 1 (LRP1; Muramatsu, 2000; Chen et al., 2007; Sakamoto et al., 2011), receptor protein tyrosine phosphatase (PTPR) Z1 (Maeda et al., 1999) and integrins (Muramatsu et al., 2004), have been identified, but there is little solid evidence indicating the involvement of a particular receptor in the signalling pathway downstream of midkine. Recently, it was reported that the activated form of Notch2 (cleaved intracellular domain of Notch2 in nucleus) was significantly decreased in the precancerous lesions of Mdk-knockout MYCN Tg mice (Kishida et al., 2013). Consistent with this, the expression of hairy and enhancer of split 1 (Hes1), a major target gene of Notch family members including Notch2, was also diminished (Kishida et al., 2013). Taken together, these in vivo results suggest that a midkine-Notch2-HES1 signalling pathway is involved in neuroblastoma. The same pathway was previously implicated in pancreatic ductal adenocarcinoma (PDAC) (Güngör et al., 2011). According to that report, the exposure of some PDAC cell lines to gemcitabine, the front-line chemotherapy used in PDAC treatment, induced both epithelial-mesenchymal transition (EMT) and a chemoresistant phenotype. It is not clear whether these two phenotypes are correlated. Although it was shown that the midkine-Notch2 pathway is involved in chemoresistance to gemcitabine, as yet, no evidence has been obtained in vivo.

Targeting midkine in neuroblastoma therapy

As listed in Table 1, several tools have been utilized to target midkine in cancer therapy. Intratumour injection of RNAi-based nucleotides [oligodeoxynucleotide (Takei et al., 2001; Takei et al., 2002), morpholino oligomer (Takei et al., 2005) and siRNA (Takei et al., 2006)] effectively suppressed the growth of xenografted tumours derived from CMT-93 cells (mouse rectal carcinoma) (Takei et al., 2001; Takei et al., 2002) and PC-3 cells (prostate carcinoma) (Takei et al., 2005; Takei et al., 2006). Intratumour injection of siRNA together with an antitumour drug also showed therapeutic effects (Lorente et al., 2011). Furthermore, i.v. injections of these oligonucleotides into an in situ model of human hepatocellular carcinoma in mice suppressed the tumours to some extent (Dai et al., 2007a,b; 2009). Also, the midkine-binding peptide has been utilized to neutralize the secreted midkine protein (Chen et al., 2007). This peptide, designated as MK-TRAP, was derived from LRP1, one of the midkine receptors. MK-TRAP consists of 169 amino acids corresponding to a part of the huge extracellular domain of LRP1, and was revealed to have a strong affinity for midkine. MK-TRAP specifically interacted with midkine, and its expression in G401 (Wilms' tumour) cells resulted in growth inhibition in the monolayer culture. In addition, not only MK-TRAP expression but also the addition of MK-TRAP-containing medium suppressed the anchorage-independent colony formation of G401 and CMT-93 cells. Furthermore, CMT-93 cell-derived xenograft tumours that express MK-TRAP have been shown to exhibit suppressed growth in vivo. In the same study, polyclonal anti-midkine antibody was also shown to exert suppressive effects on anchorage-independent colony formation with CMT-93 and G401 cells (Chen et al., 2007). Although therapeutic experiments with MK-TRAP and polyclonal antibody have not yet been done, these appear be potent tools for targeting midkine. With regard to antibodies, a recent report indicated that the growth of xenografted 143B cells (osteosarcoma) was inhibited by an i.p. injection of monoclonal antibody (Sueyoshi et al., 2012). Although this in vivo effect was not outstanding, the fact that systemic administration of this monoclonal antibody evoked a slight but clear effect suggests that it has potential for clinical application in the future. However, the use of an effective monoclonal antibody against midkine for neuroblastoma therapy will have to await further studies.

As described above, the involvement of midkine in the tumourigenesis of in MYCN Tg mice (Kishida et al., 2013) strongly indicates that midkine-targeted therapy could provide an effective treatment for neuroblastoma. In the same report, an RNA aptamer against midkine notably showed a therapeutic effect in a xenografted tumour model (Kishida et al., 2013). The RNA aptamer is the third candidate among the tools to target midkine (Table 1). RNA aptamers, which specifically recognize and directly bind to particular proteins, are considered nucleic acid analogues to antibodies. An RNA aptamer that specifically recognizes a particular molecule can be identified by screening from a complex library of random RNA sequences (20–80-mer), of typically 1014 different molecules, through a process known as systematic evolution of ligands by exponential enrichment. They form a particular three-dimensional structure and recognize a specific target in the same way as antibodies (Miyakawa et al., 2006; 2008; Ishiguro et al., 2011). Intratumour injection of the RNA aptamer against midkine (100 μg per shot, two shots per week) into neuroblastoma cell-derived xenografted tumours has been shown to suppress their growth significantly (Kishida et al., 2013). The volume and weight of the treated tumours were around a quarter of the values in control tumours (Kishida et al., 2013). These therapeutic effects of RNA aptamers were superior to those of the other tools listed in Table 1. In addition, both the activation of Notch2 and the expression of its target gene Hes1 were simultaneously attenuated in the RNA aptamer-treated tumours (Kishida et al., 2013). These results indicate that midkine-targeted therapy can be effective against neuroblastoma and also the suitability of RNA aptamer as a tool. Furthermore, the finding that the RNA aptamer treatment suppressed both the tumour growth and the Notct2-HES1 signalling pathway supports the hypothesis that the midkine-Notch2-HES1 pathway promotes the tumourigenesis of neuroblastoma in vivo. However, in contrast, it was reported that the expression level of Hes1 in neuroblastoma cell lines was quite low compared with that in T-cell acute lymphoblastic leukaemia cell lines in which canonical Notch signalling was highly activated (Zage et al., 2011). Although the expression of Hes1 is decreased concomitantly with the inactivation of Notch2 in RNA-aptamer-treated tumour cells, the significance of this change has not yet been investigated (Kishida et al., 2013). It is also possible that the canonical Delta/Jagged-Notch signal mainly induces Hes1 as a target gene, and the midkine-Notch2 signal activates the expression of other genes involved in tumourigenesis (Figure 3). This possibility should be investigated further.

On considering all the tools for targeting midkine (Table 1), the use of an RNA aptamer might be slightly preferable in terms of effectiveness, relatively low immune stimulant activity, good productivity and applicability of modifications. However, further development of much more effective monoclonal antibodies is also a promising approach that should be pursued. With regard to siRNA and shRNA, a method to accurately and efficiently deliver these to tumour cells is a big problem that needs to be solved.

Induction of chemoresistance as a function of midkine

Although it has been suggested that the midkine-Notch2-HES1 signalling pathway promotes the tumourigenesis of neuroblastoma, the final output of midkine signalling has yet to be revealed not only in neuroblastoma but also in other cancers. Recently, several reports suggested that the induction of chemoresistance was one of the functions of midkine in some cancers. Firstly, as already mentioned, midkine was shown to confer gemcitabine resistance on PDAC cells (Güngör et al., 2011). The chemoresistant PDAC cells induced midkine expression in response to gemcitabine treatment. In turn, midkine signalled via Notch2 to induce both EMT and chemoresistance against gemcitabine. Knockdown of either midkine induced by the drug or Notch2 resulted in the mesenchymal-epithelial transition and a chemosensitive phenotype. Whether MK-induced EMT can be correlated with chemoresistance has not yet been addressed. Secondly, midkine induced glioma cells to become resistant to the antitumour effects of cannabinoids (major active ingredient of marijuana) by signalling via ALK, one of the midkine receptors (Lorente et al., 2011). There are cannabinoid-sensitive and -resistant glioma cell lines, and the latter express large amounts of midkine. The targeting of midkine induced the cannabinoid-sensitive phenotype to resistant cells, while the addition of midkine made cannabinoid-sensitive cells resistant in vitro. Midkine mediated this protective effect via ALK, and the midkine-ALK signal interfered with cannabinoid-induced autophagic cell death. Furthermore, siRNA-mediated silencing of midkine sensitized the xenografted chemoresistant cells to the antitumour effects of the cannabinoid in vivo (Lorente et al., 2011) (Table 1). Other reports have also suggested that midkine induces chemoresistance in different cancers and to different drugs (Rebbaa et al., 2001; Kang et al., 2004; Mirkin et al., 2005; Hu et al., 2010; Xu et al., 2012).

In contrast to the above results obtained in PDAC cells (Güngör et al., 2011) and glioma cells (Lorente et al., 2011), all tested neuroblastoma cell lines have been found to express a certain amount of midkine (Figure 1). This is probably because the neuroblastoma cell line was mainly established from patients with poor prognosis whose expression levels of midkine were expected to be high. As the neuroblastoma cells in patients with poor prognosis express high levels of midkine, it is likely they are chemoresistant. In fact, severe chemoresistance is one of the characteristics of relapsed neuroblastoma cells, and is clearly associated with patient mortality. Unexpectedly, in contrast to PDAC cells and glioma cells, the targeting of midkine alone was sufficient to suppress the growth of these cells under a monolayer culture condition (Figure 2). PDAC cells expressed midkine only in the presence of gemcitabine (Güngör et al., 2011), and the depletion of midkine alone was not sufficient to kill chemoresistant glioma cells (Lorente et al., 2011). In neuroblastoma cells, regardless of the status of the MYCN gene (amplified in NB39 and SK-N-BE, single copy in SH-SY5Y) and the expression level of midkine (relatively low in SK-N-BE), the depletion of midkine caused significant growth inhibition in all cell lines (Figure 2). These results indicate the marked dependence of neuroblastoma cells on midkine, and highlight the strong potential of midkine-targeted neuroblastoma therapy. However, as there is still a possibility that midkine is also involved in chemoresistance in neuroblastoma cells, it would be worth evaluating the effect of combining midkine-targeted therapy with a low dose of chemotherapeutic agent (with little side effects) to achieve more effective therapeutic results. In fact, in a xenograft model of PC-3 cells, treatment with the combination of a sufficiently low dose of paclitaxel to avoid adverse effects and siRNA against midkine achieved a potent antitumour effect (Takei et al., 2006). A combination therapy targeting midkine and an antitumour drug would thus be a promising approach to clinical treatment.

Figure 1.

Midkine expression in neuroblastoma cell lines. Eight neuroblastoma cell lines were examined for the level of midkine secreted into medium by Western blotting. Each cell line in 6-well dishes was exposed to serum-free medium containing 20 μg·mL−1 of heparin sodium salt so that the heparin-bound midkine remains in the medium. Twelve hours later, the media containing heparin-bound midkine were harvested and, simultaneously, whole cell extracts were also prepared. β-actin in whole cell extracts was detected as a control for cell numbers. Six of the cell lines (SK-N-BE, NB39, IMR32, YT-nu, TGW, and TNB1) possessed an amplification of the MYCN gene, and two (SK-N-AS and SH-SY5Y) did not. Although the level of midkine expression in SK-N-BE cells was lower than that in the other cell lines, every cell line expressed a certain amount of midkine.

Figure 2.

Knockdown of midkine suppressed the growth of neuroblastoma cell lines. Three neuroblastoma cell lines (NB39, SK-N-BE, and SH-SY5Y) were infected with lentivirus to express midkine-targeting shRNA [#1 (Sigma, St Louis, MO, USA: TRCN0000303918) and #2 (Sigma: TRCN0000331252)] or non-targeting shRNA (control). In order to confirm the knockdown efficiency, the levels of secreted midkine were examined by Western blotting (each right panel). Infected cells were plated in a 96-well dish (10 000 cells/well, 3% FBS), and assayed with a Cell Counting Kit-8 (Dojindo, Kumamoto, Japan) at 0, 3, 5 day respectively. *: P < 0.001 (unpaired t-test). ALK gene statuses in each cell line are as follows: amplified in NB39 cells, normal in SK-N-BE cells, and F1174L mutation in SH-SY5Y cells.

Figure 3.

Midkine (MK) signalling in neuroblastoma cells. Neuroblastoma cells express and secrete midkine, which probably acts in an autocrine or paracrine manner. As shown here, siRNA and shRNA target midkine at the mRNA level to suppress its protein synthesis and secretion. In contrast, the RNA aptamer, antibody and MK-TRAP target the secreted midkine protein to suppress its binding to cell surface receptors. Notch2 was reported as a functional receptor of midkine in neuroblastoma cells. Although MK-Notch2 signalling induced the expression of Hes1, a canonical target gene of Notch family members, there might be other MK-Notch2-specific target genes involved in the tumourigenesis and tumour growth of neuroblastomas. Because an RNA aptamer against MK was effective in TNB1 cells that possess a hyperactive R1275Q mutation in ALK, ALK-independent MK signalling – that is, the MK-Notch2 axis – could function there. But there still might be a possibility that MK also signals via ALK in neuroblastoma cells.

Midkine receptors in neuroblastoma cells

The results depicted in Figure 2 gave us additional information about the downstream signalling of midkine. As stated previously, there are several candidates for functional receptors of midkine (Maeda et al., 1999; Muramatsu, 2000; Stoica et al., 2002; Muramatsu et al., 2004; Chen et al., 2007; Huang et al., 2008; Güngör et al., 2011; Sakamoto et al., 2011; Kishida et al., 2013). Among them, Notch2 and ALK have so far been shown to be involved in neuroblastoma.

Originally the results from in vitro experiments suggested that ALK could function as a receptor of midkine (Stoica et al., 2002). Furthermore, independently of its relationship with midkine, ALK has been shown to be a predisposition gene for familial neuroblastoma (Chen et al., 2008; George et al., 2008; Janoueix-Lerosey et al., 2008; Mossé et al., 2008). Some hyperactive point mutations within its kinase domain (e.g. F1174L and R1275Q) have been identified in familial neuroblastoma, and they were also found in some sporadic cases (somatic mutations). In addition, the development of a mouse model of this disease revealed that those ALK mutations could be a trigger for oncogenesis of neuroblastoma (Berry et al., 2012; Heukamp et al., 2012). However, the midkine-ALK axis has also been shown to be involved in the proliferation of normal sympathetic neurons in cultures of proliferating immature neurons from embryonic chick sympathetic ganglia (SG) (Reiff et al., 2011). The in vitro proliferation of primary cultured embryonic chick SG was enhanced by the expression of ALK (F1174L), and was inhibited by an ALK inhibitor or the shRNA for ALK. Similarly, the shRNA for midkine was also able to inhibit the proliferation of these cells. Importantly, although treatment with the recombinant midkine protein successfully restored the proliferation of midkine shRNA-transfected SG, it showed no effect on ALK shRNA-transfected SG. Furthermore, the growth inhibition induced by midkine shRNA was restored by the expression of ALK (F1174L). These results suggest that the midkine-ALK axis regulates the proliferation of the normal SG during embryonic development. That the midkine-ALK axis probably functions in the normal development of sympathetic neurons is an important consideration when examining the molecular mechanism of the action of midkine in neuroblastomas. This is because the normal development of sympathetic neurons and the tumourigenesis of neuroblastomas could share some molecular mechanisms. The neural crest contains several lineages of cells differentiating into melanocytes, enteric neurons, sensory neurons and so on (Nakagawara and Ohira, 2004). Among them, only the cells differentiating into sympathetic neurons have the potential to proceed to neuroblastomas. This fact suggests that the mechanism by which neural crest cells differentiate into sympathetic neurons is essential for the tumourigenesis of neuroblastoma, and that a particular abnormal event during that process should turn the tumourigenic switch on. Taken together with the results from embryonic chick SG, which suggested the midkine-ALK axis is involved in the normal development of sympathetic neurons (Reiff et al., 2011), these findings indicate that ALK functions as a receptor of midkine in neuroblastoma cells. Now, what do the results shown in Figure 2 imply? The ALK gene statuses in each cell line used in Figure 2 are as follows: amplified in NB39 cells, normal in SK-N-BE cells and F1174L mutation in SH-SY5Y cells. Hence, the results in Figure 2 indicate that the knockdown of midkine could suppress the growth of neuroblastoma cells regardless of the hyperactivated (amplification or hyperactive mutation) ALK statuses. These suggest that midkine has an ALK-independent function in neuroblastoma cells. Notch2 might mediate this ALK-independent midkine signalling (Figure 3). Of course, the possibility that midkine also signals through ALK in neuroblastoma cells cannot be ignored. The individual contributions of ALK, Notch2 and other receptors in neuroblastoma cells should be addressed in order to understand the actions of midkine.

Just as a reference, we would like to mention what is known about midkine and ALK orthologues. So far, there has been no physiological evidence to establish a ligand-receptor relationship between midkine and ALK. In order to address this point, knowledge based on the genetics of lower model animals (e.g. invertebrates) sometimes provides highly important information. That is, when a hierarchical ligand-receptor relationship is identified by the genetic study of lower model animals, the corresponding axis is frequently conserved in each mammalian orthologous gene. In terms of midkine signalling, a study in Drosophila could be interesting. The orthologues of midkine and pleiotrophin are miple1 and miple2 (Englund et al., 2006). The Drosophila orthologue of ALK has a proven ligand, named jelly belly (Jeb). Unfortunately, a mammalian orthologue of Jeb has not been identified. In addition, miple1 and miple2 show no homology to Jeb, and whether Drosophila ALK functions as a receptor of miple1 and miple2 has not been examined. Further progress in a Drosophila study would provide important information for our neuroblastoma research.

Future topics

Neuroblastoma is one of the most malignant tumours occurring in children. The current therapy consists of strong chemotherapy, radiation and surgery. Because there is almost no room to improve on these methods themselves, the introduction of a fourth method, a molecular-targeted therapy, is anticipated. Although midkine is a potent candidate for such therapy, there are several points that remain to be addressed. Firstly, the tools used to target midkine need to be refined. A significant therapeutic effect was recently achieved using intratumoural administration of an RNA aptamer (Kishida et al., 2013). However, before this tool can be used clinically, it will need to be further developed to allow for systemic administration. Because it has been suggested that some modifications could give aptamers enough stability for systemic administration (Ishiguro et al., 2011), further progress is expected. Secondly, the details of midkine signalling in neuroblastoma cells remain to be clarified. Although a growing body of evidence indicates the involvement of midkine in various cancers, the molecular mechanism by which midkine signals are transmitted in cancer cells is not fully understood. This is partly because the receptors engaged in midkine signalling are themselves uncertain. Current data suggest that Notch2, instead of ALK, would act as a midkine receptor in neuroblastoma cells (Figure 3). But it is still possible that receptors other than Notch2 and ALK function in this fashion or that multiple receptors act cooperatively to transmit midkine signalling. Solving these problems to develop a robust therapy targeting midkine could prove greatly advantageous in the fight against neuroblastoma as well as in the treatment of various other midkine-expressing cancers.

Acknowledgements

This work was supported in part by a Grant-in-Aid for Cancer Research (20-13) from the Ministry of Health, Labour and Welfare (MHLW), Japan to K.K., a Grant-in-Aid from the National Cancer Center Research and Development Fund (22-4) to K.K., Grants-in-Aid (23110002, 23390078) from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan to K.K., and Grants-in-Aid (22790311, 24590377) from MEXT to S.K.

Conflict of interest

The authors declare that they have no potential conflicts of interest to disclose.

Ancillary