An atypical tumour suppressor gene: DCC

Among the putative tumour suppressors studied, deleted in colorectal carcinoma (DCC) has been a matter of debate for years. Indeed, LOH of chromosome 18q has been found in 70% of CRC.6 ,7 This discovery led to the cloning of DCC, a transmembrane receptor bearing structural similarities with neural cell adhesion molecule protein (NCAM) family members.8 DCC expression is reduced in more than 50% of CRC, as well as in various cancers (for review, see ref. 9). Furthermore, loss of DCC is associated with poor prognosis, decreased response to chemotherapy and increased risk of metastasis.10^–17 In addition, initial data showed that restoration of DCC expression suppresses tumorigenic growth properties in vitro or in nude mice.18 Thus, the DCC gene has been proposed to be a tumour suppressor gene. Nevertheless, controversial data have rapidly emerged to counter this statement. First, few somatic mutations were initially reported for DCC in CRC and despite the fact that various mechanisms can silence a tumour suppressor, somatic mutations in cancers were, 20 years ago, commonly considered as a prerequisite for a candidate tumour suppressor. Second, other putative tumour suppressors were identified in the region of allelic loss on chromosome 18q in CRC, namely SMAD2 and SMAD4.7 Finally, mice carrying heterozygous inactivating mutations in the murine DCC ortholog did not show any cancer predisposition phenotype. These mice were also crossed with APC^min mutant mice, predisposed for CRC, in order to determine whether DCC loss might be involved in CRC progression, but the study concluded that it was not the case, closing temporally the debate on DCC’s putative tumour suppressor role.19

The debate has been reopened when DCC was shown to act as a ‘dependence receptor’.20 Such receptors are not inactive in absence of their ligand, but rather induce apoptosis, thus conferring a state of cellular dependence on ligand availability for survival (for review, see refs. 21 ,22; figure 1 and table 1). This unique functionality may confer to DCC a role of conditional tumour suppressor: it may eliminate only cells that would grow abnormally in an environment offering a limited quantity of ligand (figure 2A). In support of this view, overexpression of the DCC ligand netrin-1 in the digestive tract of mice has been shown to result in the inhibition of epithelial cell death and the promotion of tumour progression.23 It has been known for long that apoptosis regulates intestinal and colon epithelium homeostasis. Epithelial cells emerge from the stem cells niche located in the crypt and progress towards the top of the villus that they reach in 5 days. The general view is that, once arrived at the top of the villus, they undergo apoptosis and are released in the lumen. Apoptosis is also occurring when cells are damaged or stressed.24 Interestingly, netrin-1 expression is concentrated in the crypts and DCC is expressed all along the villus (figure 2B). A tempting explanation for netrin-1 overexpressing mice phenotype could be the abrogation of DCC apoptotic activity and subsequent tumour progression.23

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

The ‘dependence receptor’ concept. Schematic representation of the ‘dependence receptor’ concept. In the presence of its ligand, a dependence receptor ensures its ‘classical’ function, inducing a positive signalling leading to cell survival, differentiation, neurite outgrowth. In the absence of ligand, it actively triggers a proapoptotic signal.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

The dependence receptors, new antitumoral sentinels. (A) Dependence receptors constitute a new antitumoral mechanism, eliminating undesirable cells in a tissue where the ligand is present in a limited amount. Two selective advantages may be acquired by tumoral cells to bypass this surveillance: (i) the autocrine production of ligand (ii) or the invalidation of the receptor or of a proapoptotic partner. (B) The dependence receptor deleted in colorectal carcinoma (DCC), a new sentinel of intestine homeostasis. Intestinal epithelial cells migrate from the cryptic stem cell niche towards the top of the villus in 5 days (red arrow). They express DCC, which is bound to netrin-1 at the bottom of the villus where netrin-1 is expressed in a gradient. At the top of the villus, netrin-1 concentration is very low and unbound DCC may contribute to the apoptosis observed at the top of the villus. This apoptosis precedes the release of dead epithelial cells in the lumen. In addition, DCC may constrain the adequate number of cells all along the villus, depending on netrin-1 amount.

The proapoptotic signalling induced by unbound DCC requires its intracellular domain cleavage by caspase after aspartic acid 129025 (figure 3A). Point mutation of the aspartic acid residue (D1290N) is sufficient to abrogate DCC apoptotic activity without interfering with DCC positive signalling in the presence of netrin-1.20 Mice bearing this D1290N mutation have been recently generated. In these mice, DCC proapoptotic activity is genetically silenced.26 These mice are viable and their intestines epithelia are not disorganised, yet they display less residual apoptotic cells. Of interest, these mice spontaneously develop intestinal and colorectal neoplasia at a relatively low frequency. Loss of DCC-induced apoptosis is also associated with an increase in the number and aggressiveness of intestinal tumours in a predisposing APC mutant context, resulting in the development of highly invasive adenocarcinomas. Together with an elegant study performed by the group of A. Berns using DCC conditional mice that showed the promotion of mammary tumour metastasis in a DCC mutant background,27 this work brought the final point to the demonstration that DCC, as a dependence receptor, acts as a conditional tumour suppressor and seems to represent an important safeguard mechanism, limiting tumour progression by engaging an apoptotic process.26

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Proapoptotic signalling induced by dependence receptors. (A) In the presence of netrin-1, deleted in colorectal carcinoma (DCC) dimerises and interacts with procaspase-3. In the absence of ligand, DCC is monomeric and cleaved by caspase-3 or another protease. This cleavage exposes the addiction dependence domain to the interaction with DIP13α and caspase-9. This interaction activates caspase-9 that in turn activates caspase-3. (B) In the presence of netrin-1, UNC5H2 dimerises and its intracellular domain adopts a ‘closed’ conformation, in which the ZU5 and the death-domain interact. Death-associated protein kinase 1(DAPK) interacts with this ‘closed’ domain but stays in an inactive state via its autophosphorylation. In the absence of ligand, UNC5H2 is monomeric, its intracellular domain is cleaved and this event triggers a conformational change (‘open’ confirmation that triggers DAPK activation and a proapoptotic signalling). (C) In the absence of netrin-1, UNC5H4 is monomeric, its intracellular domain is cleaved and generates a fragment that translocates to the nucleus, interacts and activates the transcription factor E2F1 and the subsequent transcription of apoptotic genes. (D) In the absence of neurotrophin-3 (NT-3), TRKC is double-cleaved by caspase in two sites: D495 and D642. The released fragment (killer fragment (KF) 496–642) interacts with Cobra-1 and translocates to the mitochondria, where it activates Bax and induces the mitochondrial outer membrane Permeabilisation (MOMP), the cytochrome c release and the subsequent activation of apoptosome. (E) In absence of Sonic Hedgehog (SHH), Patched interacts with a complex formed of down-regulated in rhabdomyosarcoma LIM domain protein (Dral), function to Find Domain" (FIIND)-containing proteins CARD8/Nod-like receptor (TUCAN) and the E3 ligase neural precursor cell expressed developmentally downregulated protein 4 (NEDD4) that activates caspase 9 via its polyubiquitination.

Netrin-1 receptors: a new class of tumour suppressors

Interestingly, netrin-1 binds to DCC and to the uncoordinated 5 (UNC5) receptors, UNC5H1, H2, H3, H4 in rodents and UNC5A, B, C and D in humans. They are involved in neurodevelopment as well as DCC and are expressed in various adult tissues, including the intestines.28 They were also demonstrated to act as dependence receptors28–30 (table 1). The signalling pathway triggered by the UNC5 dependence receptors is still not clearly characterised, yet some pieces of the puzzle have recently emerged (figure 3B, C). UNC5 receptors are cleaved by caspase.29 ,31 ,32 UNC5H2/B and 4/D are regulated by the tumour suppressor p53. They mediate p53-induced apoptosis, a process that can be reversed by the binding of netrin-1 to its receptors.32 ,33 UNC5H2/B induces apoptosis in the absence of netrin-1 by the activation of the serine-threonine protein kinase DAPk via a PP2A-mediated dephosphorylation30 ,34 (figure 3B). UNC5H4/D's caspase cleavage releases an intracellular fragment that translocates into the nucleus and interacts with the transcription factor E2F1 to selectively transactivate proapoptotic target genes32 (figure 3C). Further studies will be required to determine whether the four UNC5 receptors trigger similar signalling mechanisms, when deprived of netrin-1.

The expression of UNC5 receptors is downregulated in various cancers, including CRC.35 The silencing of UNC5A, UNC5B and UNC5C genes appears to be mainly due to promoter methylation.35 ,36 UNC5D/H4 was identified later and is mapped in a chromosomic region frequently aberrant in many cancers.37 More precisely, UNC5D has been identified in a genome-wide study aimed at identifying genetic loci predisposing to colon carcinogenesis in mice.38 The formal demonstration that UNC5 receptors behave as bona fide tumour suppressors was provided by analysis of the mice strain with rostral cerebellum malformation (RCM) that have been shown to display a natural loss-of-function mutation in UNC5H3. These mice go through adulthood and display no obvious intestinal tumour development compared with control mice, hence suggesting that UNC5H3 inactivation in mice is not a sufficient genetic event to trigger tumour initiation. Nevertheless, when crossed with APC mutant mice, the UNC5H3^rcm/rcm mutants display an overall frequency of neoplastic lesions similar to APC^+/1638N mice, but yet show a dramatic shift of low-grade adenomas towards higher-grade tumours. Thus, UNC5H3 inactivation is associated with enhanced tumour stage in the presence of an APC mutation, demonstrating that it confers a selective advantage to CRC tumours.35

Of interest, it was reported germline mutations in UNC5C gene that may increase predisposition to CRC. Genomic DNA from human blood samples of unrelated individuals with family histories of CRC was analysed for specific UNC5C gene mutations. It led to the detection of several missense mutations in UNC5C, one of which leading to significant loss of proapoptotic activity of UNC5C and co-segregating with the disease.39 However, even though one study reported a significant increase in cancer risk in familial CRC, another recent study reports a non-significant increase in cancer risk in both familial and sporadic CRC that rather suggests that the presence of UNC5C variants in families is not sufficient to confer a high risk for CRC.40 Further investigations will be needed to determine whether this search for UNC5C variant could be used as a diagnosis marker for familial or sporadic forms of CRC.39 ,40

In conclusion, the work performed on netrin-1 receptors and their functionality as dependence receptors has allowed the identification of a new class of tumour suppressors. UNC5 and DCC receptors constitute new sentinels of the colon epithelium, eliminating cells that would overgrow in an environment presenting a limited amount of ligand. Cancer cells that are able to bypass this surveillance mechanism by silencing of the receptors or by inactivating the UNC5/DCC apoptotic activity pathway are expected to acquire a strong growth selective advantage.

Inflammation as a mechanism to inhibit netrin-1 dependence receptors-induced apoptosis in CRC

The loss of DCC and UNC5 receptors has been detected in 90% of CRC tumour samples, suggesting that it is the main selective advantage for switching off the netrin-1 dependence receptor pathway. In the remaining 10%, no loss of DCC or UNC5 was detected but rather an upregulation of netrin-1 was observed41 (table 1). These tumours appeared to display inflammation marks and increased activation of the nuclear factor-κB (NF-κB) pathway. More interestingly, when focusing on CRC associated with chronic inflammation, particularly IBD, netrin-1 upregulation was detected in more than 70% of the CRC. The mechanisms that link these chronic inflammatory states to CRC development are in large part unknown, but the main non-immune link known so far is the activation of NF-κB.42 It was indeed proposed that NF-κB pathway activation observed during IBD might contribute to tumour formation by providing antiapoptotic survival signals to the epithelial cells. Paradisi and colleagues further showed that the netrin-1 gene is a direct transcriptional target of NF-κB and that NF-κB-driven netrin-1 upregulation is associated with tumour formation in a mouse model of inflammation-induced CRC41 (figure 4). Along this line, treatment of mice with a netrin-1 titrating recombinant protein (ie, TRAP) prevents the progression of CRC developed in the classic azoxymethane–dextran sodium sulfate (AOM/DSS) inflammation-induced CRC.43 This work established the animal proof of concept that netrin-1 titrating agents could therapeutically target inflammation-driven CRC.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

The role of nuclear factor-κB (NF-κB) and netrin-1 in colorectal cancer development. Inflammation induces NF-κB nuclear translocation, via IKKβ induced IκB degradation. NF-κB activates netrin-1 transcription. Autocrine production of netrin-1 promotes tumour development.

The hedgehog dependence receptors as original regulator of CRC

The hedgehog (Hh) pathway plays many roles as a morphogen during the regulation of self-renewal and terminal differentiation in embryonic development, yet it is typically silenced in adult tissues. During gut development, Hh proteins are secreted by the epithelial cells throughout the gut and are believed to function in the underlying mesenchyme to activate molecules important in mesodermal differentiation.44

The Hh cues—Sonic Hedgehog (SHH), Desert Hedgehog and Indian Hedgehog—are believed to operate mainly via the activation of a so-called canonical pathway implying the interaction with the 12 transmembrane receptor Patched (Ptc), the loss of the repression by Ptc of an orphan seven transmembrane protein Smoothened (SMO) that leads to the downstream activation of a genetic programme via the activation of transcription factors Gli, that is, Gli1, 2, 3 (figure 5A). It was, however, reported that Ptc is also active in the absence of SHH, leading to apoptosis induction and thus functioning as a dependence receptor both in cell culture and during neural tube development45 (table 1, figure 5). Ptc, similarly to DCC, is cleaved by caspase and associates with down-regulated in rhabdomyosarcoma LIM domain protein (DRAL) and “Function to Find Domain” (FIIND)-containing proteins CARD8/Nod-like receptor (TUCAN/NLRP1) adaptor proteins, which link Ptc to caspase-9.46 Caspase-9 is then activated by polyubiquitination mediated by the E3 ligase neural precursor cell expressed developmentally downregulated protein 4 (NEDD4)47 (figures 3E and 5B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

SHH signalling pathway. (A) SHH binds to Ptc and represses SMO, allowing the activation of GLI factors. GLIA induces the expression of GLI1 and of the antiapoptotic BCL2, leading to the inhibition of apoptosis. In addition, SHH blocks Ptc cleavage, inhibiting its proapoptotic activity. (B) In the absence of SHH, Ptc inhibits SMO that triggers the degradation or the cleavage of GLI in their repressive form. GLI3R represses the expression of GLI1 and BCL2 and seems to induce apoptosis. In parallel, patched interacts with a complex formed of Dral, TUCAN and the E3 ligase NEDD4 that activates caspase 9 via its polyubiquitination. (C) HIPP1, GAS1, glypican, CDON and BOC are alternative receptors for SHH. BOC and CDON share similar extracellular motifs like the fibronectin domains type III involved in the interaction with SHH.

SHH, Sonic hedgehog; Ptc, patched; SMO, smoothened; Dral, down-regulated in rhabdomyosarcoma LIM domain protein; TUCAN, “Function to Find Domain” (FIIND)-containing proteins CARD8/Nod-like receptor; NEDD4, neural precursor cell expressed developmentally downregulated protein 4; GAS1, growth-arrest-specific 1; CDON, cell adhesion-associated oncogene regulated; BOC, Brother of CDON; GFR, glomerular filtration rate; EGF, epidermal growth factor; SSD, sterol-sensing domain; GPI, glycosylphosphatidylinositol.

Moreover, while Ptc is believed to be the main receptor for SHH, it was recently shown that SHH was also able to interact with three additional receptors, growth-arrest-specific 1 (GAS1), cell adhesion associated oncogene regulated (CDON) and Brother of CDON (BOC) that appear to contribute to the Ptc-Smo-Gli canonical pathway48 ,49 (figure 5C). Recently, it was shown that CDON, which resembles in many aspects to DCC, is also a dependence receptor50 (table 1). CDON, similarly to DCC and Ptc, is cleaved by caspase in its intracellular domain, a cleavage that allows further caspase-9 interaction and activation.50

Ptc has been described as a tumour suppressor in many pathologies51 ,52 and several loss-of-function mutations of Ptc have been reported including in CRC.52 However, so far the tumour suppressor activity of Ptc is mainly considered as a result of the fact that its loss of function activates the oncogenic-canonical Smo–Gli pathway.51 It is not known whether the dependence receptor activity of Ptc is also providing a tumour suppressive function to Ptc and even less if it is the case in CRC. The situation is definitely different for the other SHH dependence receptor. It was indeed shown that the expression of CDON is silenced in a large fraction of CRC and its expression is significantly inversely correlated with tumour grade. Moreover, CDON invalidation was shown to promote CRC tumour progression in mice because of a deficit of apoptosis in the intestinal epithelium.50 Thus, CDON behaves like a colorectal tumour suppressor.

Nevertheless, half of the CRC tumours screened in this previously cited study did not lose CDON expression and were rather shown to upregulate SHH. This phenomenon had been previously observed in various cancers.53 In cancers of the GI tract, SHH is often upregulated, thereby contributing to tumour formation and progression.44 This upregulation of SHH is believed to promote tumour progression by constitutively activating the canonical Ptc–SMO–Gli pathway. However, so far the clinical data obtained on CRC using inhibitors of the canonical pathway and showing no benefit do not support this hypothesis. One alternative hypothesis is that the gain of Hh in the tumour would not solely activate the canonical signalling (if it does so), but also or rather would block CDON and possibly Ptc-mediated tumour cell death. Along this line, it is intriguing to see the parallel between netrin-1 upregulation in CRC and NF-κB described above and the work by Kasperczyk et al54 that have demonstrated that SHH expression is under the direct control of the transcription factor NF-κB, which could establish a link between inflammation-driven cancers and SHH overexpression. They have also shown that NF-κB activation allows tumour growth and resistance to cell death via SHH upregulation.54 It was recently demonstrated, even though not directly tested in CRC models, that SHH expression in SHH-high cancer cells is a selective survival advantage that prevents CDON-induced apoptosis.50

Tyrosine kinase dependence receptors: when oncogenes turn to tumour suppressors

Tyrosine kinase receptors (TKR) are central regulators of signalling pathways that control differentiation, proliferation, motility and invasion.55 They have thus been particularly studied in the frame of oncogenic high-throughput sequencing screens. One of them concentrated on mutations altering the kinase domain of TKRs in CRC.56 This screen had been performed in the aim of finding new oncogenes and resulted in the identification of various mutations in seven TKR. Among them, five mutations were found in TRKC, also named NTRK3. TRKC, like many TKR, has long been considered as a protooncogene. TRKC and its ligand neurotrophin-3 (NT-3) have indeed been proposed to play oncogenic roles in various cancers (for review, see ref. 57). Along this line, TRKC kinase domain was found fused to ETS translocation variant 6 (ETV6) in leukaemia, lymphoma, fibrosarcoma and breast carcinoma leading to its constitutive kinase activation and thus promoting tumour formation and progression (for review, see ref. 57). Paradoxically, TRKC was also identified in a genome-wide screen for hypermethylated genes in CRC.58 This screen was dedicated at identifying tumour suppressors silenced in CRC via hypermethylation. Moreover, TRKC expression has been associated with a favourable outcome in two paediatric neoplasia, medulloblastoma59 and neuroblastoma,60 supporting the view that expression of TRKC is rather a constraint for tumour progression. TRKC could then act both as an oncogene and as a tumour suppressor? Two independent studies have possibly solved this apparent paradox in demonstrating that TRKC is a dependence receptor61 ,62 (table 1). TRKC induces apoptosis in the absence of NT-3, is double-cleaved by caspase and the released fragment is translocated at the mitochondria, via a protein named ‘Cobra1’ (Cofactor of BRCA1). This translocation allows the activation of Bax, the mitochondrial outer membrane permeabilisation, the subsequent release of cytochrome c and the apoptosome activation63 (figure 3D). As such, TRKC expression would represent a constraint for tumour escape by inducing apoptosis in settings of NT-3 limitation. Of interest, W. Grady’s team and our laboratory recently reported that TRKC expression is downregulated in CRC tumours. The silencing of TRKC is due to the methylation and deacetylation of histones of its promoter.58 ,64 Moreover, TRKC, when re-expressed in CRC cell lines, induces apoptosis and inhibits tumour progression in vitro and in vivo, supporting the view that TRKC behaves as a tumour suppressor in CRC. Of interest, one of the mutations that have been identified in the kinome screen performed by Bardelli and colleagues and that was consequently considered as a gain-of-kinase function,56 L760I, is actually affecting TRKC proapoptotic activity and thus is rather a ‘loss-of-proapoptotic-function’ mutation.58 Thus, somatic mutations affecting a tyrosine kinase dependence receptors could trigger its protooncogenic function by modulating the kinase ‘positive’ signal, as generally believed and constitute a ‘loss-of-function’ mutation altering the proapoptotic signalling. This observation brings a new light for the analysis of genome-wide screens aimed at identifying oncogenes or tumour suppressors.

Another tyrosine kinase dependence receptor has been characterised in a screen aimed at identifying hypermethylated putative tumour suppressors in CRC: REarranged during Transfection (RET)65 (table 1). RET is a well-known oncogene involved in neuroendocrine tumours especially in thyroid.66 RET, as well as TRKC, is also double-cleaved by caspase in absence of its ligand glial-derived neurotrophic factor and the released fragment has a strong proapoptotic activity.67 ,68 Very similarly to what has been recently described with TRKC, Luo et al65 brought the demonstration that RET is a bona fide tumour suppressor in CRC and is epigenetically downregulated in tumours.69

MNNG HOS transforming gene (MET) is also a well-studied oncogene that turned to be a tyrosine kinase dependence receptor, also double-cleaved by caspase70 (table 1). The behaviour of the released proapoptotic fragment resembles the one of TRKC: it translocates to the mitochondria and activates the intrinsic apoptotic pathway.71 The role of MET in CRC has long been investigated (for review, see ref. 72). Nevertheless, unlike TRKC and RET, MET is not silenced but rather upregulated in CRC in correlation with its ligand hepatocyte growth factor (HGF).73 When considering what has been described with netrin-1, SHH and the tumour suppressive function of TRKC and RET, one may wonder whether the upregulation of HGF in CRC is not only a mechanism to activate the oncogenic kinase activity of MET but also to block MET-induced apoptosis. This, however, has to be investigated further.

The observation that TKR can act as dependence receptors and, as such, behave as conditional tumour suppressors is putatively of interest in a therapeutic perspective. Indeed, tyrosine kinase inhibitors have been a major focus of the pharmaceutical industry, representing up to one-third of the drugs in development for the treatment of cancer. Yet, with the exception of some important successes in the treatment of proliferative cancers, these drugs face tumour resistances after weeks, months or years of treatment.74 The most current resistance mechanism is (i) a mutation in the kinase domain that blocks the access of the inhibitor to the ATP binding site or (ii) the increased copy number of the oncogenic kinase. One possible explanation is that inhibiting the kinase activity in cancer cells may block their proliferation but will not sufficiently kill them to avoid adaptation and clonal selection. As an example, the kinase inhibitor targeting MET (tivantinib) faces resistances in CRC, whereas antibodies preventing the binding of HGF to MET (rilotumumab, onartuzumab) are currently in phase II clinical trial and give promising results.72 Indeed, in agreement to our hypothesis, antibodies would have both the ability to inhibit the kinase activity and to enhance MET-induced apoptosis. This is why we propose that combining kinase/cytostatic drugs with agents triggering dependence receptor proapoptotic activity might bypass the kinase inhibitors-associated resistance. Thus, such agent would probably have a synergistic activity with tyrosine kinase inhibitors currently in clinical trials in the treatment of CRC, like imatinib (Gleevec), sunitinib75 and regorafenib, which all target at least the dependence receptor RET.76–78

Therapeutic perspectives linked to dependence receptors

More generally, an important question is to understand how the unexpected behaviour of receptors able to trigger apoptosis in settings of ligand limitation can be used in a therapeutic perspective. As illustrated above, various studies on several dependence receptors have clearly shown that these receptors behave as safeguards: in a tissue displaying a limited quantity of ligand, when a cell acquires oncogenic properties and starts to proliferate abnormally, it will be eliminated by its unbound dependence receptor.79 As a consequence, it is expected that cancer cells necessarily have to turn off the death signalling induced by dependence receptors. Three kinds of alterations may confer this selective advantage: (i) first, the cancer cell may silence or loose of the dependence receptor itself, as mentioned previously for DCC, UNC5, CDON, RET and TRKC in CRC. (ii) Second, the cancer cell can silence or lose a downstream proapoptotic partner of the receptor. As an example, death-associated protein kinase 1(figure 3B), a proapoptotic partner of UNC5B, is silenced by hypermethylation of its promoter in CRC.80 (iii) Third, an autocrine production of ligand can be selected by the tumour, that is, the cancer cell or the associated stroma, and this would keep the receptor fooled in an occupied state (figure 2A). The third selective advantage, that is, the autocrine production of the ligand, is probably of much larger interest for a therapeutic perspective. It has indeed been shown that biologics interfering with netrin-1/receptors interaction are able to kill netrin-1-expressing cancer cells in vitro: they display tumour growth and metastasis inhibitory effects in various preclinical models.81^–84 Accordingly, preclinical development of an anti-netrin-1 humanised monoclonal antibody is in progress. Even though CRCs will probably not be the most obvious indication as the loss of netrin-1 receptors appears to be the main selective advantage here, targeting netrin-1/receptors interaction may be of great interest in inflammatory-associated CRC.41 ,43 More generally, several studies have now demonstrated that in cancers overexpressing ligands of dependence receptors, that is, SHH in lung cancers, NT-3 in neuroblastoma, titration of the ligand is associated with tumour cell death in vitro and tumour growth inhibition in vivo.50 ,85 ,86 Whether these strategies may be useful in a fraction of CRC is of major interest. SHH as an example is known to be upregulated in CRC, and the generally accepted view is that SHH is here to activate in a paracrine or autocrine manner the Hh canonical pathway. However, so far the clinical data obtained on CRC using SMO inhibitors do not support this as they did not show any benefit, even when the patients were stratified on Hh expression.87 The view that we proposed recently was that Hh expression is actually not, or not solely, a mechanism to activate the canonical signalling, but also to block the death induced by the dependence receptor CDON.50 Along this line we provided the animal proof of concept that interfering with SHH/CDON interaction is associated with tumour cell death and tumour growth inhibition.50 As also described before, we also propose that rather than—or in association with—inhibitors of receptor tyrosine kinase, it would be interesting to develop drugs that enhance tyrosine kinase dependence receptors-induced apoptosis. Such drugs could be antibodies blocking ligand/receptors interaction. This is why the next more challenging steps of the research performed on the dependence receptor paradigm are now the future clinical work that shall assay whether interference with netrin-1/receptors, SHH/CDON or NT-3/TRKC are relevant strategies for the fraction of CRC showing upregulation of dependence receptors ligands (table 1).