The IGF system establishes the conversion of IGF ligand (IGF-I and IGF-II) activity at the cell surface into changes in cell biology. Aspects of progress in this research area have been reviewed recently by others, and here we summarize key evidence relevant to the extra-cellular components and their modification in cancer 1–7. This pathway depends on molecular interactions arranged both in series and in parallel. Release of the ligand from hydrophobic sites on inhibitory proteins precedes binding to receptors that have established competence for signalling across the membrane. Efficient and appropriate transduction of ligand binding to enzymatic amplification via phosphorylation cascades then occurs, in preference to direct protein–protein interactions. Such a system also has negative regulators that serve to dampen and modulate the amplified signals. Pleiotropic effects on parallel signalling networks may also convert a specific signalling interaction into a promiscuous activation or inhibition. Aside from immediate actions on preformed intra-cellular protein complexes controlling cellular metabolism, signalling also results in modification of nucleic acid metabolism, so closing an elaborate feedback loop. The subtle orchestration of these events is well controlled during development and is disrupted in tumours 8.
The roles of the insulin-like growth factors differ from that of insulin because, unlike insulin, their main function is not principally involved with cellular metabolism. However, the insulin and IGF signalling pathways are remarkably similar, the main differences between these two ligands being their structure, their receptors, and ligand binding preferences (Figure 1) 9. Signalling specificity may be accounted for by subtle ligand binding interactions, such as the relative duration of ligand binding to both the insulin receptor (IR) and the IGF-I receptor (IGF-IR). Further differential effects may be mediated by the cytoplasmic domains of IR and IGF-IR, and perhaps more significantly, by the quantitative differences in the relative expression of both receptors on different cell types. For example, IR appears to predominate on the cell surface of liver, muscle, and adipose tissue, compared with IGF-IR. In addition, hybrid heterodimers commonly occur between IGF-IR and IR. These hybrid receptors appear to have a slightly lower affinity for insulin than complete homodimers of the insulin receptor 9.
Both IGF-I and IGF-II are expressed during embryogenesis in the mouse, but after birth, IGF-II levels rapidly fall systemically and only remain expressed in the exchange tissues of the brain during adult life 10. In humans, probably because of differences in the gene promoter structure between humans and rodents, IGF-II is expressed at high levels in embryos, but continues to be expressed and secreted by the liver after birth and is detectable in serum throughout adult life. IGF-I is also mainly produced by the liver and increases in concentration systemically, in parallel with growth hormone-mediated postnatal and adolescent growth, before declining with increasing elderly age. Both growth factors can act close to the locality of expression (autocrine and paracrine), as well as systemically (endocrine). Excessive supply of either growth factor can induce metabolic changes similar to insulin, eg hypoglycaemia, and consequently, supply of free ligand is highly regulated.
Insulin-like growth factor binding proteins (IGFBPs) act to modulate IGF half-life, and bind with varying relative affinity to IGF-I and IGF-II. IGFBP-3 is the most abundant circulatory form, and is normally in a ternary complex with IGFs and ALS (acid-labile subunit) which retains complexes within the vascular compartment. IGFBP-5 and -6 possess at least a ten-fold higher affinity for IGF-II than for IGF-I, and the rest bind both ligands with relatively equal affinity. None of the binding proteins specifically bind insulin. All IGFBPs are substrates for a number of proteases which cleave N- and C-terminal domains of the proteins to modify ligand affinity. An additional IGF receptor, the insulin-like growth factor-II receptor (also termed the IGF-II/mannose 6-phosphate receptor), binds both IGF-II and mannosylated proteins (such as lysosomal proteases) at neutral pH with similar affinity to IGFBPs, and internalizes them for degradation within the pre-lysosomal compartment. This receptor is expressed on most cell types and acts as a specific negative regulator of IGF-II ligand bio-availability.
Through the activity of IGFs, PI3K, and PKB, phosphorylation of Forkhead transcription factors (FOXO) results in their nuclear export, and modification of the expression of a number of key genes including p27kip1, Bim, cyclin B, Gadd45, and MnSOD 11–13 (see Figure 2). Loss of FOXO function is associated with decreased expression of genes controlling cell proliferation and differentiation, and subsequent associations with transformation, eg translocations and inactivation of one allele in human leukaemia and alveolar rhabdomyosarcoma 12. Recently, integration of the Smad and IkB signalling pathways has been demonstrated to modify FOXO DNA interaction and phosphorylation, respectively 14, 15.
Insulin-like growth factor ligands, receptors, and binding protein structure
The IGFs are small, single-chain polypeptide ligands (7–8 kD) that are derived from pre-propeptides in a similar way to insulin, but contain the C-peptide bridge between B- and A-chains that is normally cleaved in insulin (Figure 3). They are also characterized by hydrophobic sites localized on the molecule surface, that render them prone both to self-aggregation at neutral pH and to binding to complementary hydrophobic sites present on receptors and binding proteins. Both IGF-I and IGF-II bind signalling receptors, IGF-IR, IR (isoforms A and B), and IGFBPs with high affinity (IGFBP 1–6). Additionally, IGF-II specifically binds to the IGF-II receptor (IGF-IIR) with high affinity. Recent co-crystallization experiments have defined the interactions between the IGFBP-5 N-terminal domain with IGF-I 16. Both the C-terminal and the N-terminal domains of IGFBPs are required for high-affinity ligand binding, with reductions in affinity by at least 100-fold when proteolytic cleavage occurs.
The structures of the IGF-IR extra-cellular domain and cytoplasmic tyrosine kinase (TK) domains have been solved by X-ray crystallography 17. Both IGF-IR and IR are composed of two extra-cellular ‘α’ subunits that bind ligand, and two trans-membrane catalytic ‘β’ subunits. The extra-cellular domain of IGF-IR has approximately 50% overall sequence identity with IR, and is composed of six structural domains which form a binding pocket. Two β-helices (L1 and L2) are separated by a cysteine-rich (CR) region and are connected to a series of three fibronectin type III domains (Fn 1–3). The L1 and CR domains appear to interact with insulin and IGF-I, respectively, but recombinant fragments of receptor L1-CR-L2 fail to bind ligand, except when combined with a short fragment of Fn1. The central Fn2 domain has a 135-amino acid sequence containing the site of cleavage for α and β subunits, and is also the site of insertion of 12 amino acids that are coded for by exon 11 of the similarly structured IR. In addition, this region also has a ligand interacting domain (including F692, F695, L696, I700, and F701), indicating that ligand binding is confined to hydrophobic regions of the α subunit 18. The cytoplasmic TK domain has 84% sequence identity between IGF-IR and IR, and on ligand binding, trans-autophosphorylation occurs by conformational change in the preformed receptor dimers at the cell surface 19.
Three tyrosine residues in the activation loop (A-loop) are phosphorylated and result in activation of catalytic activity 17. Initial phosphorylation of the activation loop at Y1135 is followed by stabilization of this structure by phosphorylation of Y1131 and Y1136, resulting in increased catalytic activity due to a major conformational change opening a gap between the NT (N-terminal) and CT (C-terminal) lobes, which allows ATP and the substrate to bind. Structural conservation of the ATP and ligand binding sites between IGF-IR and IR appears very high compared with the rest of the receptor.
Both the intra-cellular juxta-membrane and the C-terminal domains of the IGF-IR and IR component possess less sequence identity than the TK domain, and so may confer specificity. Signalling depends on phosphorylation of intra-cellular substrates, with insulin receptor substrates 1–4 (IRS1–4) and Shc (Src homology collagen) proteins being major substrates (Figure 2). Phosphorylated IRS recruits PI3 kinase, resulting in activation of phosphoinositide-dependent kinases (PDKs), protein kinase B (Akt/PKB), and p70S6 kinase. High-stringency PKB substrates include forkhead transcription factors (FOXO), glycogen synthase kinase 3β (GSK3β), mouse double minute 2 (MDM2), and the mammalian target of rapamycin (mTOR) 20. These pathways mediate gene expression, cell survival, and growth signals. As well as phosphorylation by PKB, mTOR, an atypical serine/threonine kinase, is regulated by the availability of nutrients (leucine) and regulated translation via phosphorylation of S6K (for 5′-oligopyrimidine translation) and eIF4E binding protein-1 (4E-BP1, CAP-dependent translation) 5. A further pathway is mediated through SH2 binding of the adapter Grb2 to the receptor, and leads to activation of the guanine nucleotide exchange factor SOS, Ras, Raf, and mitogen activated protein kinase (MAPK). Negative regulators also bind IGF-IR and IR TK and CT domains, and these include the Grb7/10/14 family of PH (plekstrin homology) and SH2-containing proteins 21, 22.
As a result of alternative splicing, two isoforms of the insulin receptor, A and B, have been described. As a result of expression of exon 11, IR isoform B contains a 12-amino acid insert in the Fn2 domain which reduces the affinity of the receptor for the ligands. Lack of this insert can stimulate signalling by IGF-II and appears to enhance its affinity for the IR isoform A 23.
Mutagenesis studies of IGFs and receptors, including alanine scanning, have defined the amino acids important for receptor and binding protein affinity and specificity. Similar amino acids account for the interactions between IGFBPs and IGF-IIR with IGF-II, suggesting that they may compete for ligand binding, whereas IGF-IR signalling appears to occur independently of these residues (Figure 2).
The IGFBP structure is conserved between sub-types, with IGF binding domains detected in both the N- and the C-terminal domains, with a central domain being the site of proteolytic cleavage. The C-terminal domain of IGFBP-3 and -5 also has binding sites for heparin, ALS, and a nuclear localization signal 24. These additional binding sites modulate the IGF-dependent and -independent activities of IGFBPs by altering cell surface and extra-cellular matrix localization. Proteases which are often active in tumours and cleave specific sites separating the IGF binding domains include plasmin, thrombin, prostatic-specific antigen, matrix metalloproteinases, and pregnancy-associated plasma protein-A (PAPP-A) 24. The relative efficiency of protease activity can be either dependent or independent of whether the binding proteins are bound to IGFs. When protease cleavage sites are mutated, most IGFBPs act to sequester and inactivate IGFs and inhibit functions in cell culture-based assays. Glycosylation and phosphorylation of the cleavage sites may also modify protease binding and efficiency 25. Disruption of all the murine IGFBP genes has yet to be fully reported, but judging from the first report of the IGFBP-2 gene deletion, both organ growth enhancement and reduction are observed in the context of altered levels of other IGFBPs, suggesting a significant degree of compensation 26.
Overexpression of binding proteins rescues growth effects of IGFs in murine models, supporting the predominantly inhibitory activity of IGFBPs in normal physiology 27. Inhibition of tissue proteases, for example expression of a tissue inhibitor of matrix metalloproteinase (TIMP), increases the proportion of intact IGFBP-3 within murine hepatomas, and indirectly reduces IGF-II bio-availability, as demonstrated by reduction in tumour growth 28. However, a number of studies have shown growth enhancement following tissue-specific expression of IGFBPs (reviewed in refs 24 and 29). IGFBPs also have IGF-independent effects, mediated by other motifs on the molecule. For example, IGFBP-1 and IGFBP-2 have RGD sequences which bind integrin α5β1, and evidence suggests that this interaction mediates effects on cell motility, and perhaps also to modify growth 24.
The intact IGF-IIR also binds IGF-II with high affinity (10−10M), but this interaction is not thought to be subject to proteolytic modification. The main structural domains of IGF-IIR that account for affinity and specificity are domains 11 and 13, with key hydrophobic residues (patch1) on domain 11 accounting for the main contribution of affinity 30–32.
Insulin-like growth factor ligand, receptor, and binding protein functions
Historically important experiments clearly demonstrate that addition of excess recombinant ligand to cells in culture results in a variety of cellular responses depending on the cell type and experimental conditions. The effects of exogenous IGFs fall into four main categories: increased proliferation; increased cell survival; increased cell mass; and metabolic effects. The outcome of these experiments also depended on the ratio of receptor sub-types on the cell surface, eg the ratio of IR to IGF-IR modifies the proliferative versus metabolic effects of ligand signalling. In addition, the availability of free ligand at the level of the receptor is dependent on IGFBP availability, proteolytic state, and the number and efficiency of the cycling IGF-II receptors. The interpretation of these types of experiments in the context of normal physiological restraints is further complicated by the deliberate overexpression of recombinant cell surface receptors. The normal fine control of this signalling system is exemplified by the mechanisms that exist to limit the availability of IGF-II ligand (Table 1). Restricting allele expression through genomic imprinting, protein translation, ligand bio-availability via IGF-IIR, and IGFBPs indicates that IGF-II-mediated signalling in vivo is potent and requires multiple suppressor mechanisms.
Table 1. Mechanisms that normally limit IGF-II bio-activity and their disruption in cancer
Imprinting and differential methylation
Loss of imprinting and bi-allelic expression of IGF2 mRNA
S6K translation initiation
Activated translation and increased production of IGF-II
Big-Igf2 tumour-associated hypoglycaemia
Proteolytic cleavage and increased supply IGF-II
Mutation and increased supply IGF-II
IGF-IR and IR isoform A
Up-regulation of receptor signalling
Mutation and amplification of signalling via PKB
Overexpression and activation
Haplo-insufficiency and reduced negative regulation
Genetic disruption of mammalian ligand, receptor, and binding protein genes has quantified the contribution of these components in vivo and has provided a substantial insight into the biological contribution of these players. The most profound effects of the IGF system appear to be during embryonic development, with IGF-II growth effects occurring earlier during mouse development than IGF-I effects 10, 33. Excess supply of IGF-II is lethal as a result of embryonic overgrowth, either as a result of overexpression of a transgene or from disruption of the IGF-IIR, but not as a result of disruption of imprinting 34, 35. Disruption of the IGF-IR is also lethal in the perinatal period and results in severe growth retardation, whereas disruption of binding protein genes appears to produce mild non-lethal phenotypes 26, 36. The latter result suggests redundancy of IGFBP functions, which probably underestimates their role as negative regulators of growth, which has become more evident following further in vivo experiments 37. As a result of the embryonic growth phenotypes, the function of these genes during adult life requires conditional modification to disrupt function, and tissue-specific or conditional expression to investigate gain of function.
Insulin-like growth factor ligands, receptors, and binding proteins in cancer
The development of a tumour in vivo requires a series of genetic mutations in key processes, commonly referred to as the hallmarks of cancer. The key features are mutations leading to selective growth and proliferation of tumour cells, recruitment of blood vessels and stromal cell populations, and cellular invasion.
In the context of cancer, genetic experiments in the mouse have defined key roles of system components in vivo. Experimental overexpression of both IGF-II and IGF-IR in murine tumour models results in enhanced tumour growth and progression to invasion 38, 39. Moreover, reduction in IGF signalling appears to prevent tumour progression in a variety of tumour-susceptible mouse models, including expressing T-antigen, ApcMin/+, and Ptch−/−38, 40–43. Using the rat insulin promoter to express T-antigen in islet cells results in adenoma development. IGF-II is expressed coincidently with increased angiogenesis and growth, an effect associated with loss of imprinting of the maternal Igf2 allele, and is prevented by genetic disruption of Igf241, 42. Overexpression of IGF-IR is associated with tumour invasion and metastasis in the same model 39.
A variety of human observational studies have shown that overexpression of IGF-II, up-regulation of IGF-IR, down-regulation of IGF-IIR, and modification of IGFBPs occur in common adult and paediatric human tumours 7, 44. Lack of IGF-IR-mediated cell signalling renders cells resistant to transformation by SV40 T-antigen 45, 46. An important observation to state is that mutations of extra-cellular IGF components are rare in human cancers. Unlike other growth factor receptor systems such as c-kit, EGFR, and Her2/Neu, mutations of the IGF-IR or chromosomal amplification have not been commonly described and are rare 47. However, variation in ligand supply and in IGF-IR expression level appears to be concordant with the type and aggressiveness of common human tumours, eg prostate cancer 48. However, defining a reliable quantitative approach for classifying the extent of receptor expression may prove difficult in the context of human diagnostic studies. The same problems occur to some extent with the use of immortalized cell lines and signalling molecules to define the role of the IGF system in human tumours, as signalling networks inevitably mean that numerous associations will be reported between oncogenes, tumour suppressors, and IGF-IR-mediated signalling. For example, a priming role for IGF-IR-mediated signalling does appear evident in some experimental systems, such as synergy with the suppression of proteasome function, apoptosis induced by chemotherapy, and VEGF-mediated angiogenesis 49.
Aside from mutations of gene coding for phosphatase mutated on chromosome 10 (PTEN), Ras, and other downstream signalling molecules of the IGF–PI3K–PKB–mTOR–FOXO system, the one exception to the mutation analysis of extra-cellular components has been loss of heterozygosity (LOH) and mutations of human IGF2R. Mutations and LOH of IGF2R have been detected in hepatocellular (80%), breast (40%), micro-satellite unstable colorectal (5%), and squamous lung tumours 50–54. The consequences of loss of receptor function may be a combination of increased IGF-II availability with decreased clearance of mannosylated proteolytic enzymes implicated in tumour invasion, eg cathepsin D.
A further mechanism that leads to increased IGF-II supply comes from defective epigenetic regulation of gene expression in tumours. The effect principally affects the expression of IGF2, which is normally imprinted in 90% of humans. The relative allele expression between the paternal and maternal (imprinted and so silenced) alleles is estimated to average 20–10 : 1. Following disruption of differentially methylated regions (DMRs) between the two alleles, the maternal allele can become expressed and lead to loss of imprinting (LOI) and bi-allelic expression (relative allele expression 3 : 1 to 1 : 1). The assumption is that increased mRNA expression will result in increased IGF-II autocrine and paracine supply to cells with LOI. Association studies in colorectal cancer have shown that LOI of IGF2 occurs in both the normal colon and peripheral blood of approximately 10–15% of the normal human population, an observation purported to be independent of age 55, 56. More importantly, either by allelic RNA expression identified by informative polymorphisms or by surrogate alteration in methylation of a DMR close to the IGF2 gene, LOI is associated with an increased relative risk of developing colorectal cancer. The effect does not appear to be minor, with the relative risk (RR) at least comparable to variable penetrance colorectal cancer susceptibility mutation in mismatch repair genes (RR ∼2–5). The LOI effect is not isolated to common adult tumours, but has been described in Wilms' tumours associated with LOI over-growth syndromes such as Beckwith Weidemann syndrome.
IGF-I expression is rarely derived directly from tumours themselves, but evidence points to increased risk of susceptibility to prostate, breast, colorectal, and lung cancer in humans with systemic serum and plasma levels in the high normal range (see ref 57 for meta-analysis and ref 2 for a review). Reciprocal correlation is also seen with IGFBP-3, the main circulating binding protein. Evidence supporting the hypothesis that higher levels of IGF-I increase the development and progression of tumours comes from conditional disruption of IGF-I in murine models, and the increased risk of adenoma and carcinoma development in situations where supply of IGF-I is increased, for example in acromegaly 2, 58. A criticism has been levelled at the observational population studies, in that the speculation regarding the long-term significance of this may be premature, as inter-individual variation in IGF-I levels may account for a failure to reproduce some of these data. Increasingly, the suggestion is that systemic levels may be a marker, rather than have a true functional significance and causal effect. Even with confounding factors such as whether serum or plasma was assayed, significant odds ratios were observed for prostate cancer (1.83), colorectal cancer (1.58), and pre-menopausal breast cancer (1.93) 57. Aside from measurement of systemic ligand concentrations, several studies also report associations between polymorphisms of the IGFBP-3 promoter, IGF-I CA repeat, and IRS SNPs in case-controlled studies 59.
Insulin-like growth factor ligands, receptors, and binding proteins in therapy
Major impediments to the treatment of established cancers are the relative chemotherapy and radiation resistance of tumour cells, and the narrow therapeutic index which limits the relative beneficial effects of these treatments due to toxicity to normal tissues. The differential expression of IGF system components in tumours relative to normal tissues makes this system an attractive target for therapeutic intervention. Several experimental approaches have been taken which collectively and strongly support such therapeutic developments, including cancer cell culture, xenografts in mouse, and mouse genetic models. However, all the approaches have limitations. The interpretation of data derived from transformed cell lines in culture is complicated by the use of non-physiological conditions as described above—for example, excessive exogenous addition of ligand; overexpression of recombinant receptors; relative lack of genetic controls; and the use of immortalized cell clones poorly characterized with respect to concurrent mutations of downstream signalling components. Modification of IGF-IR signalling by the addition of neutralizing antibodies, dominant negative and soluble IGF-IR, an excess of specific binding proteins, and anti-sense RNA expression—all have been demonstrated to result in reduced tumour cell proliferation and cell survival 60, 61. Recent experiments using newly developed IGF-IR kinase inhibitors (NVP-ADW742 and NVP-AEW541) have shown similar results, with approximately 1 log difference in IC50 to the IGF-IR (∼0.1 µM) compared with IR (∼2.5 µM). The most convincing evidence has come from human multiple myeloma cell lines, where IGF-IR kinase inhibition modified the expression of a significant number of genes implicated in cell cycle and cell survival responses, principally in the context of deliberate overexpression of the IGF-IR in tumour cells 49, 62, 63. Even though single agent studies show potential therapeutic advantages, such approaches also sensitize tumour cells to apoptosis induced by chemotherapeutic agents 64. Moreover, targets of the downstream pathways such as inhibition of mTOR and PKB by rapamycin can modify tumour cell survival, growth, and chemotherapy sensitivity in vivo65. Less obvious modification of IGF signalling also comes from modifying the ligand supply, such as inhibition of binding protein proteases 28.
Human trials with candidate molecules are required before the full clinical application of therapies directed at the IGF system. A number of challenges remain that may require further investigation. The acquisition of resistance in tumours is a problem that may be encountered with IGF system therapy, as gain of function of downstream components could render cells resistant to interventions, eg PTEN mutations. Moreover, ligand and receptor specificity may need to optimized, as IGF-I signalling disruption may have consequences on growth, and IR cross-reaction may induce glucose intolerance. Although initial reports are encouraging with regard to cross-reactivity between the IGF-IR and IR, the small molecule inhibitors targeting the main ATP binding site of the cytoplasmic TK domain are actually attempting to discriminate between the two signalling systems at a site where there is the most structural homology. The specificity is thus questionable in this case, but this should not deter us from further exploitation of the differences between IGFs and insulin signalling in tumour prevention and treatment, such as the ligands themselves, the cytoplasmic domains of the receptors, the IGFBPs, IGF-IIR, and specific preferences in downstream signalling molecules.
We thank Cancer Research UK, the Royal College of Radiologists, Bristol Haematology and Oncology Centre, and Special Trustees of the Bristol Royal Infirmary for support.