3.1.2. Non-Canonical Insulin Signaling and the Role of IPGs

The role for inositols in insulin signaling has long been presumed, as early experiments showed the capacity of inositols to promote glycogen synthesis in the liver or as lipid synthesis in adipocytes. The paradigm of insulin signaling changed upon the discovery of insulin modulators that were produced upon phospholipase activity in GPIs, enhancing pyruvate dehydrogenase (PDH) activity and decreasing cAMP production [66,67]. Further research lead to the description of two types of IPGs based on their structure and activity. Type A IPGs (IPG-A) contain myo-inositol and d-glucosamine and inhibit cAMP production and AMPK activity, promoting lipogenesis. The others, named as type G IPGs (IPG-G), consist of a 3-o-methyl-d-chiro-inositol (d-pinitol) and galactosamine, promoting glycogenesis via mitochondrial PDH activation [68–70]. Larner et al. carried out isolation from beef livers and later confirmed the structure of an insulin second messenger (INS-2) with a molar ratio of 1:1 of 3-o-methyl-d-chiro-inositol (d-pinitol) and galactosamine linked by a β-1,4 bond [71] (Figure 1C). INS-2 that contains an inositol glycan structure of the so-called IPG-Ps is an allosteric modulator of PP2Cα [72], which is known to dephosphorylate and activate GS [73], PI3K [74], and inactivate AMPK [75]. INS-2 might also be present under the chelated form with Mn2<sup>+</sup>. Chelated INS-2 is an allosteric modulator of mitochondrial PDH phosphatase (PDHP) activity and promotes PDH-mediated glycogen synthesis [71]. It should be remarked that the structures of DCI-GPIs are still unknown and

may not share structural similarity with MI-GPIs and differ in terms of the axial orientation of the phosphatidyl moiety, as reported by cleavage studies with synthetic DCI-GPIs (Figure 1C) [76].

Larner and colleagues described the role of a Gq/<sup>11</sup> protein as a putative pathway of insulin signaling [77], hence linking the activity of a phospholipase that could explain the release of IPGs from GPI and explain the crucial role for IPGs in insulin signaling. However, the exact structures of circulating inositols released by insulin stimulus are still unknown. This model proposed by Larner [10] raised some questions that were later added to new uncertainties in the review by Croze and Soulage [61]. Deep revision and the current data may help address some of these uncertainties.

The less widespread Müller's theory [59,60] describes the role of IPGs in activating insulin signaling externally. This may be sustained by the observation that IPG internalization is not necessary to stimulate lipogenesis in rat epididymal adipocytes with a maximal activity of 47% of the maximum insulin response [78]. This theory is based on the existence of membrane detergent/carbonate-insoluble glycolipid-enriched raft microdomains (DIGs), which are formed by the high presence of cholesterol (hcDIG) or the low presence of cholesterol (lcDIG) in the plasma membrane. Some portion of insulin receptors seem to be associated with caveolins, mainly located in "caveolae", which are structures in hcDIGs. GPI-anchored proteins would have a natural tendency to move to lcDIGs but are retained in hcDIGs by binding to a membrane protein, presumably p115. Insulin stimulus would lead to the activation of a GPI-PLC, which would release the IPGs. These IPGs may interfere in the binding of the GPI-anchored proteins to the receptor, allowing their displacement to the lcDIGs. This would also lead to a displacement of a protein kinase, pp59Lyn, previously attached to caveolin, which would mediate tyrosine phosphorylation on IRS1 or IRS2.

This theory would involve recognizing the existence of a GPI-PLC whose gene has not been identified in mammals, in addition to assuming that cholesterol microdomains are present in all cell types where insulin activity is shown, and this does not explain why IPGs may allosterically modulate intracellular elements of the insulin pathway. While this model cannot be ruled out, this may not represent a generalized mechanism and would serve as an additional route of complementary insulin signaling, but is not strictly necessary for insulin activity, rather than describing the main mechanism of action of the IPGs.

Whether there might be different IPGs contributing to insulin signaling depending on tissue or cell type might depend on the species and tissue proportion of inositol accumulation. Insulin markedly promotes the biosynthesis of DCI-GPIs after 15 min of addition to rat fibroblasts expressing the human IR, whereas a decrease in MI-GPI content is observed after 5 min of insulin treatment, which suggests that insulin promotes epimerase activity and conversion of MI to DCI [15]. DCI-containing IPGs might be the main mediators of insulin signaling, especially those involving glycogen synthesis. MI and SI are more prominent in the brain than DCI, whereas conversion of MI to DCI is far more prominent in fat, liver, muscle, or gonadal tissues [79]. It is foreseeable that DCI-IPGs would exert more important control over insulin signaling, effectively depending on the place of action.

Other debate has been raised between the intracellular or extracellular release of IPGs. The answer implies the interplay of three different proteins. Early experiments showed that IPGs are more likely to be extracellularly formed after GPI cleaving and are later actively transported in the cell. The presence of anti-IPG antibody blocks the activation of intracellular PDH, hence presuming that binding to extracellularly-generated IPGs to the antibody prevents access to the cell interior [80]. As such, some authors have described the existence of an ATP-dependent inositol glycan transporter that is stimulated upon insulin signaling. This plasmatic membrane transporter was first discovered in hepatocytes and has been well described [81]. Thus, it is a putative IPG transporter that would support the extracellular release of IPGs.

Since IPGs are part of the polar head of GPIs, their release relies on phospholipase activity. The proposed mechanism implies an alternate pathway to tyrosine phosphorylation or IRs, with IRs also coupling to a heteromeric protein Gq and the activation of a GPI-phospholipase [10]. Both GPI-PLC and GPI-PLD have been proposed as candidates. Early experiments have determined the generation of IPGs under the activity of bacterial GPI-PLC and GPI-PLD [60,68,71,82]. However, gene encoding for a mammalian GPI-PLC has not been identified yet. Presumably, insulin mediates the generation of IPGs through a GPI-PLD, as has been described [83,84]. GPI-PLD expression is ubiquitous throughout all tissues and is especially prevalent in the liver and circulating in plasma [85]. Current studies appoint the relevance of GPI-PLD in insulin resistance. Significantly increased levels of GPI-PLD have been identified as a novel biomarker of early prediabetes in humans [86] and early stages of latent autoimmune diabetes in adults and those with type 2 diabetes [87]. It has been observed that both insulin and glucose stimulate the secretion of GPI-PLD in rat pancreatic islets [88]. GPI-PLD levels also seem to be higher in the pancreas under islet hyperactivity and lower in the liver from insulin-resistant (*ob*/*ob*) mice [88].

In relation to the above, despite the fact that there is no evidence of the mammalian gene for GPI-PLC in humans, this possibility cannot be ruled out yet due to identification of a GPI-PLC-like protein in bovine brains and rat intestines [89,90], but also to the lack of knowledge of the chemical structure of the various IPGs that can be generated in the body. These doubts are raised by the experiments carried out with synthetic IPGs. In one study, it was observed that the phosphate group that binds carbon 1 of inositol to the membrane lipid needs to be maintained after cleavage of phospholipase and forms a cyclic linkage with carbons 1 and 2 of inositol for certain synthetic IPGs that have an insulin-mimetic activity [70]. This is only achieved through the action of a PI-PLC, since the hydrolysis of the phosphate is carried out on the O− radical bound to the membrane lipid, whereas a GPI-PLC performs a cleavage on the O− radical of inositol, maintaining the phosphate group in the lipid after the release of the IPG (Figure 1C) [76]. In addition to this, synthetic DCI-GPI anchors with α(1→2) linkage of glucosamine and DCI cannot be a substrate for PI-PLC hydrolysis, but this can be mediated by GPI-PLD (Figure 1C) [76]. This suggests that possible DCI-GPIs are structurally similar to MI-GPI anchors with α(1→6) linkage of glucosamine and that MI relies on GPI-PLD activity. Fagopyritols are galactose and DCI analogs found in plants and are classified according to the binding (type A with galactose-α(1→3)-DCI linkage and type B with galactose-α(1→2)-DCI linkage). Fagopyritol B1, a galactose-α(1→2)-DCI, is a structural analog of the core of the proposed DCI-GPI anchors and has a more powerful insulin-mimetic effect than free DCI, highlighting the possible role of DCI-IPGs in insulin signaling (Figure 1C) [91].

PI-PLC cannot hydrolyze and release IPGs with cyclic phosphate when inositol carbon 2 is palmitoyl-acylated, which is often the case for non-anchored protein-free GPIs [92]. In contrast, GPI-PLD may be cleaved when inositol groups are acylated, supposedly releasing acylated IPGs (A-IPGs) (Figure 1C) [92]. Synthetic A-IPGs also show a strong insulin-mimetic activity [93]. Non-protein linked GPIs are intermediate GPIs, as they quickly bind proteins when reaching the plasma membrane surface. However, non-protein-linked GPIs have been observed to reside both in the inner and outer leaflets of the plasma membrane [94]. GPI-PLD release of acylated IPGs (A-IPGs) is speculated to occur in the intracellular compartment, yet this has not been corroborated [94]. The fact that anti-IPGs block some of the insulin-mediated actions suggests that intracellularly-released A-IPGs have a minor, yet complementary, role in inulin signaling [80]. Besides insulin, GPI-PLD expression is associated with lipid levels [95–97] and its activity is also associated with triglyceride [98] and lipid metabolism in the liver [99], which may be somehow related to the improved lipid profiles of patients suffering from metabolic diseases after supplementation with inositols like DCI [100].

Assuming this approach, inositol supplementation may restore pathologically low levels of IPGs, given that the rate-limiting aspect of GPI synthesis is cytosolic-free inositol supply, as phosphoinositol synthase has a relatively low affinity [57]. Thus, a higher concentration of inositols, especially DCI, which is much more scarce than MI and has a more prominent role in insulin signaling, supports the idea that inositol supplementation would help the synthesis of DCI-GPIs and later form DCI-IPGs when insulin epimerase's activity is diminished in insulin resistance.

Albeit that this model seems to be a fairly close approach to the true role for inositols as insulin-sensitizers (Figure 2), data on supplementation with inositol derivatives might question whether IPGs are the only way for inositols to modulate insulin signaling.

**Figure 2.** Proposed mechanism of action of inositols in insulin signaling. Non-canonical insulin signaling through the Gq protein presumably stimulates glycophosphatidylinositol phospholipase D (GPI-PLD) and/or (glycol)phosphatidylinositol phospholipase C (PI-PLC), mediating the hydrolysis of phosphate linkage between inositol and membrane lipids, leading to the release of inositol phosphoglycans (IPGs). Acylated-IPGs are formed in the plasma membrane and endoplasmic reticulum. IPGs are internalized via an IPG transporter. Insulin-sensitizing properties of inositols correspond to (**A**) allosteric modulation of protein phosphatase 1A (PP2Cα) and (**B**) pyruvate dehydrogenase phosphatase (PDHP), as observed by d-chiro-inositol glycan (INS-2), and (**C**) the inhibition of adenylate cyclase (AC) and protein kinase A (PKA) activity (observed with myo-inositol glycans). IPGs may also upregulate IRS1 signaling by (**D**) activating the upstream modulator pp59Lyn localized in lipid rafts. Free inositols such as d-pinitol also exert (**E**) insulin-mimetic properties in the absence of an insulin stimulus.
