The WNT signalling pathway and diabetes mellitus

The WNT signalling pathway (referred to as the WNT pathway hereafter) was initially recognised in colon cancer research and in embryonic development studies of the fruit fly (Drosophila melanogaster), frog (Xenopus laevis) and other organisms [1, 2]. Aberrant activation of the WNT pathway may lead to the development of colorectal and other types of tumours [3]. The major effector of the canonical WNT pathway is the bipartite transcription factor β-catenin/T cell factor (β-cat/TCF). This is formed by the heterodimerisation of free β-cat with one of the four members of the TCF family (TCF7 [also known as TCF-1], lymphoid enhancer-binding factor 1 [LEF-1], TCF7L1 [also known as TCF-3] and TCF7L2 [also known as TCF-4]), TCF7L2 being the major partner of β-cat in the intestinal epithelia [3, 4]. As shown in Fig. 1a, in the absence of WNT, the cellular concentration of free β-cat is tightly controlled by a ‘destructive complex’, consisting of the tumour suppressor adenomatous polyposis coli (APC), axin, the serine threonine kinase glycogen synthase kinase-3 (GSK-3), casein kinase-1α (CK-1α) and phosphorylated ERK (pERK) [5, 6]. This complex interacts with β-cat and phosphorylates it. The phosphorylated β-cat is then destroyed via the proteasome-mediated degradation process. The WNT ligands exert their effect via the Frizzled receptors and the LDL receptor-related protein 5/6 (LRP5/6) co-receptors (Fig. 1b). Following receptor binding, WNT signals are transmitted by the association of WNT receptors with Dishevelled protein (Dvl). This event triggers the disruption of the destructive complex, preventing the phosphorylation-dependent degradation of β-cat [4], which then enters the nucleus and interacts with a member of the TCF family to form a complex that stimulates the β-cat/TCF (or WNT) downstream target genes (Fig. 1b).

Many in vitro and in vivo examinations have shown that several components of the WNT pathway are involved in pancreatic beta cell proliferation [7‐9], normal cholesterol metabolism and glucose-induced insulin secretion [10] and the production of the incretin hormone glucagon-like peptide-1 (GLP-1) [4, 11‐13]. More recently, extensive genome-wide association (GWA) studies have identified TCF7L2 as a type 2 diabetes susceptibility gene. Of all the polymorphisms studied to date, TCF7L2 polymorphisms have been demonstrated to have by far the biggest effect on the risk of developing type 2 diabetes [14‐20]. In addition, the human LRP5 gene was mapped to within the IDDM4 region, which is linked to type 1 diabetes on chromosome 11q13 [21‐23]. Polymorphisms in LRP5 have been shown to be associated with obesity phenotypes, and missense mutations in LRP6 have been shown to be associated with the risk of bone loss, early coronary disease and the metabolic syndrome [24, 25]. Moreover, polymorphisms in Wnt5B have been shown to be associated with the risk of type 2 diabetes [26].

Investigations into how β-cat, the major effector of WNT signalling, influences pancreatic islet development using transgenic and knockout mice has produced inconsistent results. Although several WNT ligands and the Frizzled receptors were shown to be produced in the embryonic and postnatal pancreas [27‐29], an early study showed that the loss of β-cat did not significantly perturb pancreatic islet endocrine cell mass or function, although β-cat is essential for pancreatic exocrine acinar cell development [30]. Utilising the pancreatic and duodenal homeobox protein 1 (Pdx1)–Cre system to specifically delete the gene encoding β-cat (Ctnnb1) in the epithelium of the pancreas and duodenum revealed that Ctnnb1-deleted cells had a competitive disadvantage during pancreas development [31]. Although there was a reduction in endocrine islet numbers during early embryonic development and the mice developed pancreatitis perinatally because of the disruption of acinar epithelial structure, the mice later recovered from the pancreatitis and regenerated normal pancreas and duodenal villi from the wild-type cells that escaped the Ctnnb1 deletion [31]. However, a more recent study found that inducing the production of a stabilised form of β-cat (a mutant one that is more resistant to the proteasome-mediated degradation process) at different stages of development has different effects. During the early stage of organogenesis, robust production of stabilised β-cat drives changes in Hedgehog and fibroblast growth factor signalling and blocks the expression of the gene encoding Pdx-1, an important homeodomain protein transcription factor involved in the genesis of pancreatic beta cells. Induction of the stabilised form of β-cat at a later time point in pancreas development enhances proliferation and increases the size of this organ [32].

Rulifson et al. recently examined the effect of WNT signalling in regulating beta cell genesis and proliferation using both in vitro and in vivo approaches [7]. Purified WNT3a (which is known to activate the canonical WNT pathway) stimulated proliferation of both the mouse beta cell line MIN6 and primary mouse pancreatic beta cells, possibly through the cell cycle regulators cyclin D1, cyclin D2 and cyclin-dependent kinase 4, as well as the homeodomain transcription factor Pitx2. Immunohistological examinations of 3-month-old bi-transgenic rat insulin I promoter (RIP)-Cre and β-cat^active mice revealed a threefold increase in the production of Ki67 by pancreatic beta cells, which occurred in parallel with a 2.5-fold increase in beta cell mass [7]. Furthermore, axin production led to impaired Pitx2 gene expression, along with impaired beta cell expansion [7]. Taken together, these observations suggest that WNT signalling is necessary and sufficient for pancreatic beta cell proliferation.

The incretin hormone GLP-1 has been shown to stimulate WNT activity in adult mouse pancreatic islets, and both TCF7L2 and β-cat are required for GLP-1-stimulated proliferation of the rat pancreatic beta cell line INS-1 [9]. Furthermore, adipocyte-derived WNT molecules have been reported to induce beta cell proliferation and insulin secretion in vitro [33].

Besides the effectors of WNT signalling, the co-receptors of the WNT ligands, LRP-5/6 [34, 35], are also important for normal lipid and glucose metabolism. An early study demonstrated that loss of function mutations of LRP5 were associated with the development of the autosomal recessive disorder osteoporosis-pseudoglioma syndrome [36]. The co-receptor was shown to be important in transducing WNT signalling and to play critical roles in modulating bone accrual and eye development [37]. The human LRP5 gene maps within the IDDM4 region on chromosome 11q13, which is linked to type 1 diabetes [21‐23]. A recent GWA study has also shown that polymorphisms in LRP5 are associated with obesity phenotypes [38]. Pancreatic production of LRP5 has been reported [39, 40].

It has been demonstrated that LRP5 can interact with axin, one of the inhibitors of the WNT pathway (Fig. 1). When Lrp5 was expressed in fibroblast cells, the LRP5 protein alone exerted no effect on the WNT pathway, but acted synergistically with the WNT ligands [41]. Furthermore, LRP5 molecules without the extracellular domain were constitutively active. They induced TCF/LEF-mediated transcription and stabilised β-cat [41]. The addition of WNT ligands to the medium triggered the translocation of axin to the cell membrane and enhanced the interaction between axin and LRP5. Finally, the LRP5 domain involved in the interaction with axin is also required for TCF/LEF-mediated transcriptional activation. These observations collectively suggest that binding of axin by LRP5 and its translocation to the cell membrane is an important part of WNT signal activation [41] (Fig. 1b).

After being fed a high-fat diet, Lrp5 ^−/− mice showed increased plasma cholesterol levels and after being fed a normal diet they showed markedly impaired glucose tolerance [10]. Furthermore, in response to the administration of high concentrations of glucose, the mice showed significant reductions in intracellular ATP and calcium levels and decreased glucose-induced insulin secretion [10]. The WNT ligands WNT3a and WNT5a stimulated insulin secretion in the wild-type mice but not in the LRP5 ^−/− mice [10], suggesting that WNT ligands require a functional LRP5 to regulate insulin secretion.

GLP-1 is an important incretin hormone that is encoded by the Gcg gene, which is expressed in the intestinal endocrine L cells [42‐44]. In these cells, expression of Gcg mRNA and production of GLP-1 can be activated by lithium, which mimics the function of the WNT ligands [45], or the overproduction of the constitutively active S33Y β-cat mutant [12], indicating that Gcg is a downstream target of the WNT signalling pathway. Activation was subsequently attributed to a TCF binding site within the G2 enhancer element of the Gcg promoter and the production of TCF7L2 in the intestinal endocrine L cells [4]. It is well known that Gcg expression and GLP-1 production can be activated by elevations in cAMP levels [46‐52]. Since the G2 enhancer element has been shown to mediate the stimulatory effect of both cAMP and calcium on Gcg promoter activity [53], it is possible that cAMP pathway cross-talks with the WNT pathway to regulate Gcg expression [12].

Insulin inhibits Gcg expression in pancreatic alpha cells [54, 55]. This inhibition is physiologically important because Gcg mRNA expression in pancreatic islets leads to the production of glucagon, the primary counter-regulatory hormone of insulin [54, 55]. A recent study showed a significant stimulatory effect of insulin on Gcg mRNA expression and GLP-1 production in intestinal L cells. Interestingly, insulin activated a Gcg–luciferase (LUC) reporter gene construct containing the wild-type TCF binding site within the G2 enhancer element of the Gcg promoter, but not the one carrying a mutation at this site. Either ‘knocking-down’ β-cat production or the function of TCF7L2 completely blocked insulin-stimulated intestinal Gcg expression. Thus, insulin, at least partially uses effectors of the WNT pathway to exert this stimulatory effect [13]. Interestingly, the stimulatory effect of insulin on Gcg promoter activity was blocked by phosphatidylinositol 3-kinase (PI3K) inhibition, but not by protein kinase B (PKB) inhibition [13]. This, along with the observations of the existence of PKB-independent PI3K activity in other cell lineages [56‐58], suggest that an unknown signalling component mediates PI3K-mediated cross-talk between insulin and WNT signalling pathways [13]. In hyperinsulinaemic and insulin-resistant MKR mice [59, 60], Gcg mRNA expression and GLP-1 production in the distal ileum were significantly higher than in sex- and age-matched controls [13], indicating that hyperinsulinaemia and/or insulin resistance may affect the homeostasis of GLP-1 production. Figure 2 shows that both WNT and cross-talk between insulin and WNT are involved in GLP-1 production, although detailed mechanisms and the pathophysiological significance of the cross-talk need to be further explored. Figure 2 also shows that lithium and cAMP may stimulate Gcg expression by affecting the production of free β-cat.

In 2006, a study reported that inheritance of specific single nucleotide polymorphisms (SNPs) within the TCF7L2 gene was related to an increased risk of type 2 diabetes [20]. The investigators of this study have genotyped 228 microsatellite markers in Icelandic individuals with type 2 diabetes and healthy controls across a 10.5 Mb interval on chromosome 10q. A microsatellite, DG10S478, located within the intron 3 region of TCF7L2, was found to be associated with type 2 diabetes [20]. This observation was subsequently replicated in both a Danish and US cohort [20]. The investigators found that two of five SNPs investigated within introns 4 and 5 of TCF7L2, namely rs12255372 and rs7903146, were in strong linkage disequilibrium with DG10S478 and showed similarly robust associations with type 2 diabetes [20]. This discovery has drawn attention globally [15‐19, 61‐71], and studies in many other ethnic groups have confirmed that rs12255372 and rs7903146 are the two SNPs most strongly associated with type 2 diabetes [72, 73], with the SNP rs7903146 reported to have the greatest effect in white individuals [73, 74]. The two SNPs occur at relatively low frequencies in Asian populations, although an association with type 2 diabetes was identified in two large Japanese cohorts [75, 76]. Recent studies have revealed two novel SNPs associated with the risk of type 2 diabetes. The SNP rs290487 was identified in a study of a Han Chinese population in Taiwan [77], and the SNP rs11196218 was identified in a study of Hong Kong Chinese individuals [78]. It has been reported that non-diabetic carriers of the risk-associated TCF7L2 SNPs do not have defects in GLP-1 secretion [64].

Since TCF7L2 is known as an intestinal cell specific transcription factor [79] and is an important regulator of intestinal Gcg expression and GLP-1 production [11], it was suggested that the TCF7L2 SNPs may modify disease susceptibility by affecting intestinal Gcg expression and plasma levels of GLP-1 [20]. More recently, genotyping 1,100 non-diabetic German individuals for the five known TCF7L2 SNPs indicated that TCF7L2 variants are associated with reduced insulin secretion [64]. In contrast, plasma GLP-1 levels during an OGTT were not significantly influenced by the TCF7L2 variants [64]. The CT/TT genotypes of the SNP rs7903146 were shown to strongly predict future type 2 diabetes in two independent Scandinavian cohorts [63]. The risk T allele was associated with impaired insulin secretion, incretin effects, and an enhanced rate of hepatic glucose production [63]. Furthermore, investigators found that islet TCF7L2 expression was increased fivefold in individuals with the TT genotype. Although TCF7L2 expression was positively correlated with the expression of INS, which encodes insulin, it was inversely correlated with glucose-stimulated insulin release [63]. Furthermore, an ex vivo examination demonstrated that TCF7L2 knockdown (with small interfering RNA) increased human pancreatic beta cell apoptosis and reduced beta cell proliferation and glucose-stimulated insulin secretion [8]. Overexpression of TCF7L2, on the other hand, protected islets from glucose- and cytokine-induced apoptosis and from impaired functions [8]. As discussed above, both TCF7L2 and β-cat are required as effectors for GLP-1-stimulated beta cell proliferation [9].

Although these recent studies indicate that TCF7L2 SNPs may directly affect INS expression and/or insulin secretion, we still do not have a clear picture of how these SNPs affect the function of pancreatic beta cells. As discussed by Schafer and colleagues, the involvement of changes in GLP-1 production and secretion influenced by TCF7L2 variants in the increased risk of type 2 diabetes cannot be eliminated [64]. It should also be pointed out that the participants in their study were non-diabetic individuals. It is possible that a certain compensatory response(s) attenuated the defect in GLP-1 secretion in these TCF7L2 SNP carriers in the pre-diabetic stages.

Over the last few years, investigations have led to the discovery of the insulin–FOXO protein signalling cascade [80‐83]. This regulatory system controls metabolic homeostasis and other important physiological and pathophysiological events. In the absence of insulin or growth factors, FOXOs are mainly located within the nuclei and upregulate the expression of a set of target genes, thereby promoting cell cycle arrest, stress resistance and apoptosis (Fig. 3a). In the presence of insulin or growth factors, FOXOs are phosphorylated by PKB and serum- and glucocorticoid-regulated protein kinase and stay in the cell cytosol (Fig. 3b) [81]. In this way, FOXOs function to control the growth, development, metabolism, and possibly, longevity of the cell and the organism in response to insulin, insulin-like growth factor-1 and many other growth factors [81]. In contrast to the effect on insulin signalling, oxidative stress induces the activation of FOXO signalling [84]. It appears that this is due to the activation of the small GTPase Ral, which leads to the c-Jun NH₂-terminal kinase (JNK)-dependent phosphorylation of FOXOs, followed by their translocation to the nucleus and increased FOXO-mediated transcriptional activities [84].

An evolutionarily conserved interaction between the WNT pathway effector β-cat and FOXOs was discovered in 2005 [85]. In mammalian cells, a yeast two-hybrid screen detected an interaction between β-cat and FOXO1 and FOXO3. The gene bar-1 (also known as C54D1.6) encodes the Caenorhabditis elegans (nematode worm) homologue of β-cat [86], while the FOXO gene homologue in this organism is daf16 [87]. An interaction between BAR-1 and DAF16 was also detected [85]. In mammalian cells, binding of β-cat to FOXO enhanced the transcriptional activity of FOXO. In C. elegans, the loss of BAR-1 reduced the activity of DAF16 in dauer formation and life span [85]. More importantly, the association between β-cat and FOXO was shown to be enhanced in cells exposed to oxidative stress [85].

The interaction between FOXO and β-cat prompted scientists to explore the pathophysiological role of this interaction in age-dependent diseases, including osteoporosis. It was reported that male and female mice with sufficient sex hormones could still lose bone mass and strength progressively during the ages of 4–31 months, and that this was associated with enhanced osteoblast and osteocyte apoptosis, reduced osteoblast number and bone formation rate, and elevated levels of reactive oxygen species (ROS) [88]. Furthermore, in the C57BL/6 mice, ageing was shown to be associated with the reduced expression of several WNT target gene mRNAs, including Axin2 and Opg (also known as Tnfrsf11b); and increased expression of the FOXO target genes, such as Gadd45 (also known as Gadd45a) [89]. However, hydrogen peroxide treatment in an uncommitted mesenchymal cell line, C2C12, increased FOXO-mediated transcription and attenuated both basal and WNT3A-stimulated levels of Axin2 and other WNT target genes. Opposite effects of hydrogen peroxide on FOXO- and TCF-mediated transcription were confirmed by measuring the reporter gene activity of FOXO-LUC and TCF-LUC in response to different dosages of hydrogen peroxide. More recently, the concept of that the interaction of FOXO with β-cat inhibits β-cat/TCF activity was further confirmed [90]. It has been demonstrated that small interfering RNA (siRNA)-mediated knockdown of FOXO reverted the loss of β-cat binding to TCF after cellular oxidative stress [90]. The production and function of FOXO proteins in the pancreatic beta cells have been extensively investigated during the past few years [91]. It will be interesting to examine whether insulin and growth factors control the balance between FOXO-mediated and WNT-mediated gene transcription (see below).

Comprehensive in vitro and in vivo studies by multiple laboratories have shown that WNT signalling is important in normal pancreatic islet development, as well as in pancreatic beta cell function and genesis. It appears that the WNT/β-cat pathway plays direct, precise, and even opposing roles during different stages of pancreatic islet development [32]. Obviously, to exert such precise and opposing effects, WNT/β-cat signalling needs to interact with other signalling pathways and regulate the downstream gene expression profiles, both temporally and spatially.

WNT signalling is also important in activating intestinal Gcg transcription and, therefore, the production of the incretin hormone GLP-1 [11‐13], which has been shown to utilise WNT signalling effectors, i.e. β-cat/TCF7L2, to exert its effect on beta cell proliferation [9]. Further examinations are required to verify whether TCF7L2, certain WNT ligands and Frizzled receptors are also involved in the genesis of pancreatic islets and intestinal endocrine L cells.

GWA have revealed relationships between SNPs in LRP5, which encodes a co-receptor of the WNT ligands, and the risk of type 1 diabetes [22, 23], as well as obesity [29]. LRP6 mutations are possibly related to the development of bone loss, coronary disease and the metabolic syndrome [24, 25]. More importantly, extensive recent studies have identified associations between SNPs in TCF7L2 and the risk of type 2 diabetes. These observations suggest that WNT signalling is not only involved in pancreatic islet development during embryogenesis, but also in the function of pancreatic and intestinal endocrine cells during adulthood. Since all the known risk-associated SNPs of TCF7L2 are located within the intronic regions, the effect of these SNPs on TCF7L2 expression should be examined. To ultimately understand why these SNPs affect the risk of type 2 diabetes, we need to explore mechanisms underlying TCF7L2 production in pancreatic and intestinal endocrine cells under both physiological and pathological conditions.

Type 2 diabetes is a chronic and age-dependent disease. As shown in Fig. 4, nuclear FOXOs increase during ageing because of the accumulation of ROS and JNK signalling pathway activation [92]. FOXOs compete with TCF proteins, including TCF7L2, for the limited pool of β-cat. This leads to reduced WNT activity, which is important for lipid and glucose metabolism, pancreatic beta cell proliferation and function, and the production of the incretin hormone GLP-1. Insulin and growth factors, on the other hand, may restore the balance between FOXO- and TCF-mediated gene transcription by trapping FOXOs within the cell cytosol. The establishment of this concept offers a new perspective on the pathogenesis of type 2 diabetes and other age-dependent diseases. Further examination of the cross-talk between insulin/growth factors and WNT signalling pathways may lead to the development of novel therapeutic approaches for the treatment of type 2 diabetes and other age-dependent diseases.