Address correspondence and reprint requests to Dr Stephen K. Fisher University of Michigan, MHRI Laboratories at MSRBII, 1150 West Medical Center Drive, C560 MSRBII, Ann Arbor, MI 48109–0669, USA. E-mail: email@example.com
Inositol phospholipids and inositol phosphates mediate well-established functions in signal transduction and in Ca2+ homeostasis in the CNS and non-neural tissues. More recently, there has been renewed interest in other roles that both myo-inositol and its highly phosphorylated forms may play in neural function. We review evidence that myo-inositol serves as a clinically relevant osmolyte in the CNS, and that its hexakisphosphate and pyrophosphorylated derivatives may play roles in such diverse cellular functions as DNA repair, nuclear RNA export and synaptic membrane trafficking.
The prominent role of inositol phospholipids in signal transduction events within the CNS is now well established (for reviews, see Fisher and Agranoff 1987; Fisher et al. 1992). In addition, roles for phosphoinositides as facilitators of a diverse array of cellular events, such as membrane trafficking, maintenance of the actin cytoskeleton, regulation of cell death and survival, and anchors for plasma membrane proteins, have been documented (for reviews, see Toker and Cantley 1997; Low 2000; Vanhaesebroeck et al. 2001). In contrast, with the notable exception of inositol 1,4,5-trisphosphate and its pivotal role in the mobilization of intracellular calcium, less emphasis has been placed on the functional significance of the water-soluble components of the phosphoinositide pathway, i.e. myo-inositol and its phosphorylated derivatives. To date, seven of the 63 possible inositol phosphate esters have been assigned putative physiological roles (Irvine and Schell 2001), and many others have been detected in a variety of organisms. In addition, at least four pyrophosphorylated inositol phosphates and enzymes that synthesize (Huang et al. 1998; Saiardi et al. 2001b) and degrade them (Safrany et al. 1998; Caffrey et al. 2000) in brain have been identified. Given their structural diversity, it appears likely that cellular functions of other inositol phosphates remain to be discovered.
In this review, we focus on the role that myo-inositol itself and some of its highly phosphorylated derivatives (namely, inositol hexakisphosphate and the diphosphoinositol polyphosphates) may play in neural function. The CNS is an atypical tissue in that it possesses relatively high concentrations of myo-inositol as well as the means to synthesize it. myo-Inositol serves not only as a precursor molecule for inositol lipid synthesis, but also as a physiologically important osmolyte. Alterations in brain and CSF inositol concentrations have been reported in a number of pathological conditions. In addition, oral administration of inositol has been claimed to be of therapeutic benefit in the treatment of several neuropsychiatric disorders. Inositol hexakisphosphate (IP6; phytate) is the most abundant of the inositol phosphates in both neural and non-neural cells. In common with myo-inositol, IP6 is particularly prevalent in the CNS. IP6 binding sites have been identified (Sasakawa et al. 1995), and a number of functions highly relevant to the CNS have been proposed. IP6 was held to be the most highly phosphorylated inositol-containing molecule until the discovery of pyrophosphate-containing derivatives, which has added yet another level of complexity to our understanding of both the metabolism and the function of inositol phosphates. The diphosphoinositol polyphosphates are readily synthesized by enzymes derived from the CNS and exhibit a rapid turnover, and are thus likely to play significant roles in neural function.
A crystalline product derived from muscle extracts and having the empirical formula of a carbohydrate was discovered some 150 years ago (Scherer 1850) and was termed ‘inosit’, from inos the Greek root for muscle, and which was translated into English as ‘inositol’. It was later predicted to be one of nine possible stereoisomers of hexahydroxycyclohexane (Bouveault 1894), of which seven were correctly predicted to be optically inactive and of the remaining two to be an enantiomeric pair (Fig. 1). The inositol isomer described by Scherer was found to be naturally abundant, and was eventually identified as its hexakisphosphate ester, phytic acid, a common component of many foods, especially grains. To distinguish the name of the prevalent isomer in nature from the eight other isomeric inositols, it was renamed myo-inositol. When one uses the term inositol without a prefix, it is generally presumed that one is referring to the myo isomer, whereas the term ‘inositols’ refers to all nine. Inositol was one of a number of ‘bios’ growth factors established to be required by some microorganisms (Eastcott 1928). A nutritional deficiency in laboratory animals could not be easily demonstrated, presumably because of the ubiquitous presence of inositol in the diet, and the ability of some tissues to convert d-glucose-6-phosphate to d-inositol-3-phosphate, which could then be dephosphorylated to inositol. In addition, dietary phytate can in some instances give rise to free inositol, either by the action of ingested dietary phytases or from endogenous phytase activity in the intestinal mucosa. Cultured human cell lines that are dependent on inositol have been identified, some of which can synthesize inositol, but presumably not in sufficient quantities to support maximal growth (Eagle et al. 1960).
Although the human diet typically provides 1 g/day of inositol (Holub 1986), approximately 4 g/day could be synthesized by the kidneys (Clements and Diethelm 1979; see Fig. 2). Brain and testis also synthesize inositol (Hauser and Finelli 1963), but the kidney is by far the major organ involved in its catabolism and excretion. Inositol degradation occurs via a specific oxygenase that cleaves the polyol ring with the resulting production of d-glucuronic acid (Charalampous 1958; Arner et al. 2001). The latter is then metabolized in the liver to d-xylulose 5-phosphate and enters the pentose phosphate pathway (see Fig. 2).
A special function for inositol in the nervous system was first suggested by the discovery of a relatively high inositol content in the brains of many species (Young 1934; Long 1961). Folch found that cephalin contained lipid-bound inositol (Folch and Woolley 1942), which was eventually shown to be one of three phosphoinositides: phosphatidylinositol (PI), phosphatidylinositol-4-phosphate (PI4P) and phosphatidylinositol-4,5-bisphosphate [PI(4,5)P2]. Evidence that the inositol lipids are metabolically highly active was first noted in manometric studies (Sloane-Stanley 1953), and later by means of 32P labeling (Eichberg and Dawson 1965). Eventually, the high rate of turnover could be explained by evidence that they were components of the phosphoinositide-mediated signal transduction system (Berridge 1987).
Most of the other eight isomers of myo-inositol, largely the result of chemical or microbial action on myo-inositol hexaphosphate produced by plant grains, have been isolated as phosphate esters from soil. A number of derivatized cyclitols, mainly methyl ethers, are also present in plants (Posternak 1965). Scyllo-inositol is present in human brain in quantities estimated to be from 5 to 12% that of myo-inositol (Seaquist and Gruetter 1998). Rat and rabbit brains contain an epimerase that converts myo- to scyllo-inositol via a myo-scyllo-inosose intermediate (Sherman et al. 1968a,b), but it is not known whether brain scyllo-inositol is synthesized in situ or how much of it is transported into the brain via the blood (Spector 1978). Levels of both myo and scyllo isomers can be measured non-invasively in the human brain by means of 1H-spectral shift NMR. Neo-inositol has been reported to be present in mammalian brain, but at much lower concentrations (Sherman et al. 1971). Small quantities of labeled chiro-inositol are reported in the rat brain following intraperitoneal injection of 3H-labeled myo-inositol (Pak et al. 1992). At present, phosphates of inositols other than myo-inositol have not been reported in brain or other animal tissues. However, it has been recently learned that Entamoeba histolytica contains neo-inositol phosphates (Martin et al. 2000).
Distribution of inositol in CNS and other neural tissues
Inositol in tissues and other biological samples can be isolated from other polyols by TLC and, following derivatization, separated by GLC, and then identified by retention times and quantified by means of a variety of detectors (cited in Kouzuma et al. 2001). It can also be selectively measured in a mixture by means of specific bacterial dehydrogenases (Kouzuma et al. 2001). With the exceptions of the testis, kidney and lens of the eye, the mammalian CNS contains significantly higher concentrations of inositol than non-neural tissues. As is the case in non-neural cells (Sigal et al. 1993), it is likely that inositol in neural tissues is not in a single compartment (Wolfson et al. 2000). If the assumptions are made, for purposes of comparison, that each wet weight gram or mL of brain contains 100 mg of protein, and that its dry weight comprises 20% of wet wight, then intracellular concentrations of inositol present in whole brain, gross anatomical regions or peripheral nerves can be calculated to range from 2 to 15 mm. These concentrations of inositol exceed those present in the CSF and plasma by 25–50- and 50–100-fold, respectively (see Table 1). Within the CNS, relatively similar concentrations of inositol are reported for distinct regions of either human (Shimon et al. 1998) or rat brain (Patishi et al. 1996). However, at least in human brain, the absolute values may not remain constant throughout life because concentrations of inositol in temporal cortex decline by > 50% between the ages of 20 and 90 (Stokes et al. 1983). Inositol is also readily lost from brain tissue when cell integrity is compromised. Thus when slices are prepared from cerebral cortex, hippocampus or cerebellum, there is a 50–60% loss of inositol (Table 1).
Table 1. Concentrations of myo-inositol in neural tissues
aInositol intracellular concentrations are calculated on the basis that each g wet wt of tissue contains 100 mg of protein. Dry weight is assumed to comprise 20% of wet weight (Long 1961). bSee reference for additional brain regions. cCalculated on the basis of measured intracellular volumes.
It remains uncertain to what extent inositol is preferentially concentrated within neuronal or glial cells. It has been suggested that inositol is primarily (Glanville et al. 1989), or even exclusively (Brand et al. 1993), localized to the glia. However, inositol concentrations in some neuronal populations may equal or exceed those observed for glia. Concentrations of inositol approaching 10–20 mm have been reported for the Purkinje cell layer of the cat cerebellar flocculus (Godfrey et al. 1982), and for the giant neurons of Deiter's nucleus in the rabbit (Sherman et al. 1977). These values are similar to those obtained for cultured human NT2-N neurons or SH-SY5Y neuroblastoma cells (Novak et al. 1999). In other neuronal types, inositol concentrations are distinctly lower. For example, the concentration of inositol in cultured rat cerebellar granule cells is < 25% of that in NT2-N neurons (Novak et al. 1999), and HCN-2 human neurons are reported to be less than 1 mm in inositol (Koch et al. 1999).
Molecular mechanisms for maintenance of inositol concentrations
In addition to the recycling of inositol phosphate second messengers, cells maintain their intracellular concentration of inositol via three distinct mechanisms: transport of inositol across the plasma membrane via specific carrier molecules, de novo synthesis of d-inositol 3-phosphate from d-glucose 6-phosphate, and the efflux of inositol, which is mediated via a volume-sensitive organic osmolyte channel.
Concentrations of inositol in neural cells typically exceed those in plasma by a wide range, from 2- to 500-fold (Sigal et al. 1993). This concentration of the polyol is accomplished primarily by a saturable Na+-dependent uptake mechanism that exhibits a high affinity for inositol and is mediated by a specific sodium myo-inositol transporter (SMIT) molecule, originally cloned from canine MDCK cells (Kwon et al. 1992). From its cDNA, a protein of 718 amino acids with 12 membrane-spanning domains is predicted (molecular weight; MW = 79.5 kDa). Both the amino and carboxyl termini of SMIT are cytoplasmic, and several potential phosphorylation sites for either protein kinase A or protein kinase C are found in the cytoplasmic domains. Canine SMIT exhibits 40–50% homology with the Na+/d-glucose cotransporters of rabbit intestinal mucosa and kidney proximal tubules (Hediger et al. 1987; Coady et al. 1990). Human SMIT is highly homologous to the canine transporter (Berry et al. 1995). The gene encoding human SMIT (SLC5A3) appears to be a complex transcriptional unit that consists of at least five exons subject to alternate splicing (Porcellati et al. 1998), and is overexpressed in Down's syndrome (see below).
The transport characteristics of cloned renal SMIT were examined in detail following expression in Xenopus oocytes (Hager et al. 1995). Uptake is pH-dependent and requires two Na+ ions for each molecule of inositol transported. Half-maximal uptake occurs at 50 µm, and phlorizin is a potent competitive inhibitor. Myo- and scyllo-inositols are the preferred substrates for transport, with the rank order efficacyfor transport for a number of carbohydrates beingmyo-inositol = scyllo-inositol > l-fucose > l-xylose > l-glucose, d-glucose, 3-O-methyl-d-glucose, 2-deoxy-d-glucose > d-xylose. In contrast, the specificity for transport by the Na+/glucose cotransporter (SGLT1) is d-glucose, α-methyl-d-glucopyranoside, d-galactose, d-fucose, 3-O-methyl-d-glucose > d-xylose, l-xylose, 2-deoxy-d-glucose > myo-inositol, l-glucose, l-fucose. Uridine is transported by an Na+/nucleoside transporter, which exhibits considerable homology to both SGLT1 and SMIT (61% and 47%, respectively), although neither SMIT nor SGLT1 is able to transport uridine.
SMIT, as determined by expression of its mRNA, is widely distributed throughout the CNS and is found in both neural and non-neural cells. There are striking regional differences in the expression of SMIT mRNA, with highest levels observed in the choroid plexus, pineal, hippocampus, locus cœruleus and Purkinje cells. Although the distribution of mRNA for SMIT is distinctly uneven, all brain regions possess the transcript for the transporter (Inoue et al. 1996). In contrast, only relatively small differences in Vmax (0.2–0.5 nmol/mg protein/min) have been reported for SMIT measured in astrocyte preparations derived from major brain regions, including cerebral cortex, hippocampus, cerebellum, diencephalon and tegmentum (Isaacks et al. 1997, 1999c; Lubrich et al. 2000). Km values for the high affinity transport of inositol in these distinct regions are also relatively similar (13–50 µm; see Table 2). In terms of development, expression of both the mRNA and SMIT protein is highest in fetal brain and declines postnatally (Guo et al. 1997). This pattern of expression may be linked to the high concentration of inositol in fetal CSF (Battaglia et al. 1961). As observed for the adult brain, mRNA for SMIT is diffusely distributed throughout the fetal brain and spinal cord with positive signals observed for both neural and non-neural cells. The developmental pattern for SMIT in kidney is completely opposite of that observed for brain: renal SMIT mRNA expression is low during embryogenesis and increases significantly after birth (Guo et al. 1997).
Table 2. Kinetic characteristics of Na+/myo-inositol transporter in neural tissues
Both the expression and activity of SMIT can be regulated by a number of factors, the best documented of which is osmotic stress. This has been demonstrated both in vivo (Ibsen and Strange 1996) and in vitro for cultured glia (Strange et al. 1991, 1994; Isaacks et al. 1997), neuroblastoma (Wiese et al. 1996) and human NT2-N neurons (Novak et al. 1999). In response to hyperosmolar stress (> 330 mOsm), the expression of SMIT mRNA increases several-fold, as does the activity of the transporter. SMIT gene expression can be regulated by five tonicity-responsive enhancers scattered throughout the 5′-flanking region of the gene (Rim et al. 1998). The increase in rat SMIT mRNA occurs only after several hours and precedes increased activity of the transporter (Ibsen and Strange 1996). The latter, which is associated with an increase in the Vmax of SMIT without a change in Km, is blocked in the presence of cycloheximide, thereby indicating a requirement for de novo protein synthesis. Conversely, when cells are returned to iso-osmotic conditions, SMIT mRNA levels decrease rapidly and inositol is slowly lost by passive efflux. Whereas SMIT transcripts are broadly distributed throughout the brain, there are marked regional differences in the magnitude of increased expression of SMIT mRNA in response to hypertonicity (Ibsen and Strange 1996; Bitoun and Tappaz 2000). Following increases in plasma osmolarity, significant increases in SMIT mRNA levels were observed for the rat cerebral cortex, habenula and nucleus accumbens, but little increase occurred in the caudate-putamen. Similarly, plasma hypertonicity increases SMIT mRNA in the inferior, but not in the superior, colliculi (Ibsen and Strange 1996). Increases in mRNA for SMIT were also observed following the intraperitoneal administration of kainic acid, a treatment that results in an intensive neuronal discharge and a concomitant increase in the concentration of intracellular electrolytes. The increases in SMIT mRNA were most pronounced in the frontal and parietal cortex (Nonaka et al. 1999).
SMIT expression is also regulated by inositol. Thus, when cultured astrocytes obtained from the cerebellum, cerebral cortex, hippocampus or tegmentum were pre-treated with 400 µm inositol for 1–15 days there was a time-dependent reduction of SMIT activity, which reached its maximum (approximately 30%) at day 8 (Lubrich et al. 2000). In contrast, SMIT activity in astrocytes obtained from the diencephalon is not down-regulated by inositol. The latter observation, along with the differential regulation of SMIT activity by corticosterone and hyperosmolarity (Lubrich et al. 2000), may suggest the presence of distinct SMIT isoforms in different brain regions. SMIT may also be regulated by anti-bipolar therapeutic drugs such as Li+, valproate or carbamazepine, each of which is reported to reduce the activity of the transporter (Lubrich et al. 2000). The activity of SMIT may also be regulated by protein kinase C, the activation of which results in an inhibition of the transporter in human astrocytoma cells (Batty et al. 1993) and human NT2-N neurons (Novak et al. 2000). In contrast, activation of protein kinase A does not alter the activity of SMIT in NT2-N neurons (Novak et al. 2000).
A novel mammalian inositol transporter has recently been identified from the screening of databases for sequences related to SGLT. However, this molecule does not transport glucose or other hexoses, but is an H+-myo-inositol symporter (HMIT; Uldry et al. 2001). HMIT cDNA encodes a 618 amino acid polypeptide with 12 predicted transmembrane domains (MW approximately 75–90 kDa, depending upon the extent of glycosylation). HMIT is expressed almost exclusively within the CNS, primarily within astrocytes, but it is also present in some neuronal populations. Of interest are the pH-dependencies of the two transporter types. Thus whereas SMIT is progressively inhibited by a reduction in pH (Matskevitch et al. 1998), HMIT is maximally active at low pH. Although the physiological significance of HMIT has yet to be established, this transporter might mediate an additional regulatory mechanism of inositol homeostasis.
Inositol is synthesized de novo from glucose in a number of organs, including brain, testis, kidney and liver (Hauser and Finelli 1963). The trimeric enzyme d-inositol-3-phosphate synthase catalyzes the irreversible cyclic aldol condensation of glucose-6-phosphate to yield d-inositol-3-phosphate (l-inositol-1-phosphate; Eisenberg 1967; Culbertson et al. 1976; Maeda and Eisenberg 1980). The reaction requires nicotinamide adenine dinucleotide (oxidized) (NAD), is stimulated by NH4+ and K+, and is inhibited by Li+. The d-inositol-3-phosphate produced is then hydrolyzed by the Mg2+-dependent Li+-inhibitable enzyme inositol monophosphatase (IMPase). No synthase activity could be detected in human NT2-N neurons (Novak et al. 1999). This result is consistent with previous immunocytochemical data localizing the brain enzyme primarily to the capillary endothelium (Wong et al. 1987).
Putative roles for inositol in neuropathological conditions
A number of conditions have been identified in which alteration in inositol disposition may play a role, for example either as a physiologically important osmolyte or as a precursor molecule for phosphoinositide synthesis.
Disorders of volume regulation in the nervous system
In functioning neural and non-neural cells, electrolytes are transported across the plasma membrane, and the composition of osmotically active particles inside the cell is in a constant state of flux. Unless the osmolyte concentration is tightly regulated, bulk flow of water across the plasma membrane may result in damaging changes in cell volume. To maintain cell volume, acute changes in tonicity are regulated by changes in the transport of Na+, K+, H+ or Cl– across the plasma membrane, whereas chronic changes in tonicity are offset by the transport of ‘compatible’ or ‘non-perturbing’ organic osmolytes, primarily inositol (Strange 1992; Lang et al. 1998).
At least part of the pathogenesis of Down's syndrome may be related to the function of inositol as an osmolyte in the nervous system. Band q22 of human chromosome 21, which contains the gene encoding SMIT, is triplicated (Berry et al. 1995), and the inositol concentrations in the brain and CSF are 30–50% higher than in controls (Shetty et al. 1995b; Shonk and Ross 1995; Berry et al. 1999; Huang et al. 1999). Fetal CSF, which has a 10-fold higher concentration of inositol than that of the adult mammal (Battaglia et al. 1961), is even further enriched in Down's syndrome, and the excess of the polyol has been linked to osmolyte and electrolyte imbalances in the developing CNS. Inositol concentrations are also elevated in Ts65Dn mice, which possess three copies of mouse chromosome 16 – homologous to human chromosome 21 (Shetty et al. 2000). The triplication of SMIT results in the predicted 50% increase in inositol uptake in the trisomic fetal cortical neurons (Acevedo et al. 1997).
Inositol transport is also altered in stroke, a condition in which neurons around a focal ischemic zone become hypoxic, with a concomitant loss of ATP. The latter results in a rise in intracellular electrolyte concentrations and in an increase in intracellular osmolyte loads caused by the accumulation of catabolites (Venkatachalam and Buja 1994). Both of these events contribute to cytotoxic edema, which would ordinarily be countered by inositol efflux (Nonaka et al. 1998). However, because VSOAC activity requires ATP, efflux is diminished, thereby exacerbating neuronal swelling (Jackson et al. 1994). Physical trauma to the brain can compromise the blood–brain barrier, with a concomitant leakage of ions and macromolecules into the cerebral extracellular space, which in turn results in vasogenic edema and the triggering of an up-regulation of SMIT mRNA and cellular inositol uptake (Yamashita et al. 1997).
Alterations in inositol transport are also implicated in hepatic encephalopathy. This condition is associated with hyperammonemia, which results in astrocyte swelling and a loss of cerebral inositol (Haussinger 1994). In cultured astrocytes incubated with NH4+, there is a marked reduction in inositol uptake, which reflects both a reduction in the number of transporter molecules and their affinity for inositol. Ammonia also increases the fast efflux of inositol from these astrocytes (Isaacks et al. 1999b). Both the reduction in inositol influx and the accelerated efflux of inositol may represent adaptive changes to counteract the ammonia-induced swelling of astrocytes.
Bipolar disorder: Li+ and the inositol depletion hypothesis
The discovery that lithium selectively blocks the action of IMPase (Hallcher and Sherman 1980) was successfully exploited by Berridge and colleagues in concert with the use of [3H]inositol to demonstrate ligand-stimulated activation of the phosphoinositide-mediated second messenger system in intact cells (Berridge et al. 1982). They further hypothesized that the well-known psychotherapeutic action of Li+ in mania could be attributed to its block of IMPase (Berridge et al. 1989), with a concomitant reduction in the synthesis of phosphoinositides. Li+ administered in vivo leads to an accumulation of inositol monophosphate (Allison et al. 1976). When brain slices are incubated in the presence of Li+, receptor activation results in a substantial increase in the formation of inositol mono- and bisphosphates, whereas conversely, the production of inositol tris- and tetrakisphosphates is reduced (Batty and Nahorski 1985; Kennedy et al. 1989, 1990; but see also Lee et al. 1992). The depletion hypothesis proposes that Li+ also decreases free inositol that is required as cosubstrate with cytidine diphosphodiacylglycerol (CDP-DAG, see Fig. 2) for PI formation via PI synthase. Depletion of the inositol in the presence of Li+ could then slow the metabolic sequence; whereby PIP2, the precursor of the messenger molecules I(1,4,5)P3 and DAG, is regenerated. In keeping with this possibility, a reduction in PIP2 supply has been observed in a non-neural cell line (CHO) transfected with the cDNA for the M1 muscarinic receptor (CHO-M1) following chronic receptor activation in the presence of Li+ (Jenkinson et al. 1994). An appealing aspect of the inositol depletion hypothesis is the uncompetitive nature of the Li+ block of IMPase: the inhibition is enhanced by its substrate, inositol monophosphate. Thus, in ‘hyperactive’ cells, with rapid phosphoinositide recycling and a concomitantly larger production of inositol monophosphates [I(1)P, I(3)P and I(4)P], greater inhibition by Li+ is proposed to occur, with a lesser block in ‘normal’′ cells. However, in experimental animals administered Li+ at doses equivalent to human therapeutic levels, there is only a small decrease in brain inositol concentration (Allison and Stewart 1971), which remains above the established Km for PI synthase (Benjamins and Agranoff 1969; Ghalayini and Eichberg 1985; Imai and Gershengorn 1987). In contrast, the Li+ block leads to a 20-fold increase in total brain inositol phosphates, and it has been proposed that it is this increase that mediates the therapeutic action of Li+, rather than inositol depletion (Agranoff and Fisher 2001). Some neuronal populations may contain as much as 10–20 mm inositol, a finding that would seem to further lower the probability that inositol levels in such neurons can be lowered sufficiently in the presence of Li+ to slow the phosphoinositide cycle. This objection would be overcome if the inositol recycled in the phosphoinositide cycle were in an intracellular pool that excluded the bulk of brain inositol. Alternatively, Li+ may target populations of neurons in which inositol concentrations are constitutively low. Experimental demonstration of Li+-mediated accumulation of CDP-DAG, cosubstrate with inositol in the PI synthase reaction, would indicate depletion of the inositol substrate. However, for CDP-DAG to accumulate experimentally, prior depletion of cellular inositol is required, as occurs in the preparation of brain slices (Godfrey 1989; Heacock et al. 1993). In the case of cultured cells, the ability of Li+ to augment receptor-mediated increases in CDP-DAG accumulation is cell-specific. Thus, whereas Li+ strongly potentiates CDP-DAG accumulation in both cerebellar granule cells and CHO-M3 cells following carbachol addition, no such increase is observed for SH-SY5Y cells – a result attributed to the high concentration of inositol in this neuroblastoma cell line (Gray et al. 1994). A number of studies have reported a reversal of a biological action of Li+ by inositol. For example, inositol is reported to reverse teratogenic actions of Li+ in developing Xenopus embryos (Busa and Gimlich 1989). In this case, further investigations revealed the inhibition of glycogen synthase kinase by Li+, and evidence was presented to indicate that this metabolic site, rather than IMPase, accounts for the deleterious effect that Li+ has on development (Klein and Melton 1996).
Involvement of inositol in other neurological disorders
A role for inositol, either direct or indirect, has been proposed for a number of other disorders of the CNS. For example, intraperitoneal administration of inositol to curly tail mice, a genetic model of folate-unrelated neural tube defects, significantly reduced the incidence of spina bifida in fetuses (Greene and Copp 1997). The beneficial effects of inositol were mimicked by the administration of activators of protein kinase C and could be prevented by Li+. It is suggested that inositol administration results in a stimulation of phosphoinositide turnover, which in turn up-regulates the β-retinoic acid receptor, a gene that is expressed at reduced levels and linked to neural tube defects. Administration of inositol can also ameliorate the symptoms associated with experimental diabetic peripheral neuropathy. The human condition is associated with decreased Na+-K+-ATPase activity and slowing of nerve conduction velocity, both of which are reversed in fucose-fed rats upon dietary supplementation with inositol (Yorek et al. 1993). The inhibition of inositol uptake in peripheral nerves has been implicated in the etiology of the disease. The proposed pathway involves hyperglycemia-induced activation of aldose reductase, leading to sorbitol accumulation and subsequent inhibition of inositol uptake, which results in a lowered cellular inositol concentration (Gillon and Hawthorne 1983). Inositol uptake is also reduced in fibroblast cell lines obtained from patients with ataxia telangiectasia, an autosomal recessive disorder characterized, among other symptoms, by a progressive degeneration of the cerebellum (Yorek et al. 1999). Finally, although inositol concentrations are unaltered in Alzheimer's disease (Stokes and Hawthorne 1987), it has recently been shown that myo-inositol can form a stable complex with Aβ42, a soluble form of amyloid present in plaques. Epi- and scyllo-inositols (but not d-chiro-inositol) were also capable of forming complexes. The interaction of inositol with Aβ42 results in a complex that is non-toxic to both PC-12 cells and primary human neurons (McLaurin et al. 2000).
Use of inositol in the treatment of human neuropsychiatric disorders
Barkai et al. (1978) reported that significantly lower concentrations of inositol were present in the CSF of both unipolar and bipolar depressed patients than were in healthy control subjects. Some 15 years later, it was demonstrated that oral administration of high doses of inositol (6 g/day) could increase the concentration of inositol in the CSF by as much as 70% (Levine et al. 1993). This observation prompted investigations into the use of inositol in treatment of depression. Oral administration of 12 g/day of inositol reportedly resulted in a significant improvement in depression, as ranked by the Hamilton Depression Rating Scale, whereas the discontinuation of treatment with inositol led to a rapid relapse in a significant number of patients (Levine et al. 1995; Levine 1997). Inositol administration is also claimed to be effective in the treatment of both panic disorder (Benjamin et al. 1995) and obsessive compulsive disorder (Fux et al. 1996). The reported beneficial effects of inositol became evident only after 4–6 weeks of treatment, a time frame similar to that required for most therapeutic agents, including Li+. Although there is clear NMR evidence that oral administration of inositol is reported to result in a 20% transient increase in inositol concentrations within the brain (Moore et al. 1999), the mechanism underlying a possible therapeutic improvement remains obscure. An unexplored possibility is that the increased production of glucuronic acid that accompanies a large intake of inositol increases the excretion of P 450-hydroxylated xenotoxic or autotoxic metabolites as their glucuronides.
Inositol hexakisphosphate (IP6 or ‘phytic acid’) was originally identified some 80 years ago as the principal phosphate storage molecule in plants (1–3% by weight of whole grains and cereals). Subsequently, it has been determined that IP6 is ubiquitously distributed in animal cells, and is one of the most abundant of the inositol phosphates. Separation and quantitation of highly phosphorylated inositol phosphates from biological samples were initially problematical because high salt concentrations are required for elution from ion-exchange HPLC columns, which complicates the detection and analysis. Many of these problems can be circumvented by the use of the metal dye detection method (Mayr 1990). High voltage electrophoresis separation at low pH (Seiffert and Agranoff 1965) separates IP6 and IP5 isomers from one another in a few minutes and has proven useful in biological and soil samples (Cosgrove 1980). 31P NMR analysis has been employed in the identification of higher inositol phosphates (Scholz et al. 1990), and 31P/1H two-dimensional NMR has also been applied successfully to the analysis of phytate hydrolysates (Johnson and Barrientos 1995). Concentrations of IP6 in distinct brain regions range from 10 to 15 µm (Yang et al. 2001). Intracellular concentrations of IP6 are estimated to be in the range of 10–100 µm for animal cells, whereas in the slime mold Dictyostelium, it can be 10 times higher (Shears 2001). Intracellular availability of phytate is limited by the insolubility of its Ca2+/Mg2+ salts and its avid electrostatic binding to protein. Apart from some indications that the concentrations of IP6 can change during the cell cycle, there are few documented examples of situations in which the concentration of IP6 changes significantly following cell stimulation. This had led to the suggestion that the metabolic turnover of the IP6 is sluggish – certainly quite distinct from that observed for other inositol phosphates, such as I(1,4,5)P3. However, this impression may be misleading as more than one metabolic pool of IP6 may be present in cells. Moreover, as discussed below, there is strong evidence for the rapid phosphorylation of IP6 into the diphosphoinositol polyphosphates ‘IP7’ and ‘IP8’, and these products in turn can be dephosphorylated back to IP6 (see Fig. 3). Thus a portion of the intracellular IP6 may represent an intermediate, rather than an end point, in the metabolism of inositol phosphates.
The way in which IP6 is synthesized in mammalian cells is not yet established with certainty. Unlike the situation that exists in plants and Dictyostelium, in which IP6 can be synthesized from inositol (via d-inositol 3-phosphate) following a series of positionally selective, sequential kinase reactions (see Irvine and Schell 2001), there is no evidence for the existence of this pathway in mammalian cells. In the latter, it appears more likely that the synthesis of IP6 proceeds via the action of a newly discovered inositol polyphosphate multikinase (mIPMK; Saiardi et al. 2001a), which can sequentially phosphorylate I(1,4,5)P3 in both the 3′ and 6′ positions to yield I(1,3,4,5)P4 and then to yield I(1,3,4,5,6)P5 (see Fig. 3). Alternatively, I(1,3,4,5)P4 can be synthesized via the action of an I(1,4,5)P3 3-kinase, a highly specific enzyme that is enriched in neural tissues and is likely to have evolved from mIPMK (Irvine and Schell 2001). Direct evidence for the presence of an IP5 2-kinase (the enzyme that would complete the synthesis of IP6) in mammalian cells is currently lacking, although such an enzyme has been documented in yeast (Ives et al. 2000).
A diverse array of putative physiological functions has been assigned to IP6. For example, in non-neural tissues, IP6 has been suggested to be anti-neoplastic (Jariwalla 1999), to prevent kidney stone formation (Grases et al. 1998) and to act as a cellular antioxidant (Graf et al. 1987). In plants, IP6 was found to regulate stomatal pore closing (Lemtiri-Chlieh et al. 2000). Interest in the role that IP6 may play in the CNS stems from the fact that a large number of high-affinity IP6 binding sites can be detected in membrane preparations from distinct brain regions (Sasakawa et al. 1995). Autoradiographic studies have indicated that the distribution of these IP6 binding sites is consistent with their localization to neuronal cell bodies. IP6 has also been shown to bind with even higher affinity to a number of purified proteins, several of which are closely associated with membrane trafficking events. These include the vesicle adaptor proteins AP-2 (Voglmaier et al. 1992) and AP-180 (Ye et al. 1995), arrestins (Gaidarov et al. 1999) and synaptotagmin (Fukuda et al. 1994). As a consequence, roles for IP6 in receptor regulation (Sasakawa et al. 1995), vesicle trafficking (Gaidarov et al. 1996) and neurotransmitter release (Llinas et al. 1994; Ohara-Imaizumi et al. 1997) have been proposed. In addition, IP6 has been shown to bind with high affinity to myelin proteolipid protein and l-type Ca2+ channels. It has also been proposed to regulate PKC and protein phosphatases (see Shears 2001 and references therein). However, because IP6 is so highly negatively charged, it has a propensity to bind non-specifically, and thus the assignment of physiological functions to IP6 must be made with caution. In this regard, Shears (2001) has proposed a series of criteria that would distinguish IP6-specific from IP6-non-specific actions, including the ability of IP6 to elicit the response in the presence of physiologically relevant concentrations of Ca2+ and/or Mg2+, the inability of Ca2+ chelators to mimic the effect of IP6, demonstration that other inositol phosphates, especially the diphosphoinositol polyphosphates, are either inactive or exhibit a lower potency and a lesser ability of inositol hexakissulfate or other conformers of IP6 (e.g. neo- and scyllo-inositol hexaphosphates) to mimic the action of IP6. When these criteria are rigorously applied, the purported ability of IP6 to regulate a number of physiological events becomes less certain. For example, several of the assays in which high-affinity IP6 binding to proteins has been demonstrated, e.g. to myelin proteolipid protein (Yamaguchi et al. 1996), were conducted in Ca2+-free media. Furthermore, in several instances, the diphosphoinositol polyphosphates appeared to bind more avidly than IP6 to a proposed target protein, e.g. AP-180 (Ye et al. 1995). Calcium chelators have also, on occasion, been found to mimic the effect of IP6, e.g. as in a proposed extracellular neuromodulatory role for IP6 (Vallejo et al. 1987; Sun et al. 1992). Nonetheless, an IP6-dependent protein kinase that phosphorylates pacsin/syndapin I, a protein involved in synaptic vesicle recycling, has recently been identified that on the whole satisfies most of the stated criteria (Hilton et al. 2001).
Two putative roles of IP6 linked to nuclear function have emerged recently. IP6 is reported to activate a repair mechanism for radiation- and chemically induced double-stranded breaks in DNA via activation of a DNA-dependent protein kinase (Hanakahi et al. 2000). The latter, which is part of a multimeric DNA end-joining complex, possesses a putative binding domain for IP6. Other inositol phosphates, as well as inositol hexakissulfate, are less effective, and the effect of IP6 can be observed at physiological Mg2+ concentrations (Hanakahi et al. 2000). A second proposed nuclear role for IP6, based upon mutation studies in yeast, is the regulation of nuclear mRNA transport. York et al. (1999) have identified four yeast genes involved in mRNA export, three of which encode for enzymes involved, either proximally or distally, in IP6 synthesis. Deletion of IP6 kinase (see below) does not impair mRNA export, thereby eliminating the involvement of the diphosphoinositol polyphosphates (Saiardi et al. 2000). Additional support (albeit correlative) for a link between IP6 availability and mRNA export comes from studies with mammalian cells in which transfection with the SopB protein (a Salmonella virulence protein, which can act as an inositol phosphatase) results in a reduction of both parameters (Feng et al. 2001), although results obtained with SopB need to be interpreted with caution because of the multiple effects of SopB on cellular processes (see Shears 2001). These results, taken collectively, nonetheless raise the intriguing prospect that IP6 plays a key role in nuclear function.
Much of our present knowledge regarding the metabolism of these inositol polyphosphates is derived from studies with 3H- or 32P-labeled substrates and cloned recombinant enzymes. Although there existed previous evidence from anion-exchange chromatography for the presence of inositol-containing compounds that were more polar than IP6 (Europe-Finner et al. 1991), the first clear-cut identification of inositol phosphates containing energy-rich pyrophosphoryl residues was obtained in 1993 from two independent laboratories. Investigations with cell-free lysates obtained from Dictyostelium or AR4–2J pancreatoma cells that had been incubated with radiolabeled IP6 (Menniti et al. 1993; Stephens et al. 1993) or, alternatively, intact hepatocytes incubated with [3H]inositol (Glennon and Shears 1993), led to the identification of three diphosphoinositol polyphosphates: diphosphoinositol tetrakisphosphate (PP-IP4), diphosphoinositol pentakisphosphate (PP-IP5, also known as ‘IP7’) and bisdiphosphoinositol tetrakisphosphate ([PP]2-IP4, also known as ‘IP8’; see Fig. 3). The concentrations of the diphosphoinositol polyphosphates are relatively low in mammalian cells (1 µm), but are up to 300 times higher in Dictyostelium (Stephens et al. 1993). Structural analyses utilizing metal-dye-detection HPLC procedures reveal that in mammalian cell lines the major isomers are likely to be 5-PP-IP4, 5-PP-IP5 and 5,6-(PP)2-IP4 (Albert et al. 1997).
The precursor for PP-IP4 synthesis is IP(1,3,4,5,6)P5. The latter can then be phosphorylated to PP-IP4 by the action of mIPMK (Saiardi et al. 2001a). Alternatively, any one of the three identified IP6-kinase isoforms derived from distinct genes and cloned from mammals (Voglmaier et al. 1996; Saiardi et al. 2000, 2001a) can also phosphorylate IP5 to PP-IP4 (and PP-IP4 to (PP2)-IP3– see Fig. 3). However, IP6 is the preferred substrate for the enzyme. The three IP6 kinases (K1, K2 and K3; MW = 49, 49 and 46 kDa, respectively) exhibit 45–65% similarities in amino acid sequences, and all are highly enriched in brain (Saiardi et al. 2001a). The Km for IP6 is approximately 1 µm, whereas the enzyme exhibits a much lower affinity for its cosubstrate, ATP (1–2 mm). When transiently transfected in a non-neural cell, human embryonic kidney cell 293 (HEK293), the K2 and K3 isoforms are preferentially localized to the nuclei and cytoplasm, respectively, whereas the K1 isoform is found at both subcellular locations. A selective kinase that preferentially phosphorylates PP-IP5 to (PP)2-IP4 (‘IP7’ kinase) has also been purified from brain (Huang et al. 1998), but has not yet been cloned. This enzyme is enriched in the cytoplasm, and has a MW of 56 kDa and exhibits a 10-fold higher affinity for PP-IP5 than for IP6 (1 vs. 10 µm). Both IP6 and PP-IP5 kinases can act reversibly under physiological conditions and thereby act as ATP synthases. These observations have prompted the suggestion that diphosphoinositol polyphosphates may act as primary phosphate donors to acceptors other than ADP, such as proteins (Voglmaier et al. 1996; Huang et al. 1998), but none have as yet been identified.
Diphosphoinositol polyphosphates are very active metabolically. This becomes evident when their breakdown is prevented by the inclusion of fluoride to inhibit phosphatase action (Menniti et al. 1993). By use of this F–‘metabolic trap’ technique, it has been calculated that 20–50% of the cellular IP6 reservoir cycles through the diphosphoinositol polyphosphates per hour. A similar flux occurs between IP5 and PP-IP4 (Menniti et al. 1993). The observation that diphosphoinositol polyphosphates turn over rapidly has stimulated interest in their potential physiological roles. Most emphasis to date has been placed on the possibility that membrane trafficking events are regulated by these pyrophosphorylated inositol phosphates. In yeast, deletion of the IP6 kinase results in a disruption of vacuole biogenesis (Saiardi et al. 2000). In mammals, PP-IP5 binds tightly (Kd = 20 nm) to AP-3 (a clathrin adaptor protein) and is a potent inhibitor of AP-3-dependent assembly of clathrin cages (Ye et al. 1995). Clathrin is a major protein in brain (∼1% of total), and its assembly is a key event in endocytosis. PP-IP5 also binds with high affinity to AP-2, another adaptor protein that promotes the assembly ofclathrin cages (Shears et al. 1995). Because dephosphorylation of PP-IP5 to IP6 should result in a less potent inhibition of clathrin assembly, this metabolic step could conceivably result in the activation of endocytosis (Safrany et al. 1999). Of possible relevance is the recent observation that PP-IP5 (like IP6) can stimulate a protein kinase that selectively phosphorylates pacsin/syndapin I (Hilton et al. 2001), a protein involved in synaptic vesicle recycling. In addition, PP-IP5 binds with high affinity to coatomer, a cytosolic multimeric complex that also controls membrane trafficking events between the Golgi complex and endoplasmic reticulum (Fleischer et al. 1994; Ali et al. 1995).
The turnover of diphosphoinositol polyphosphates may also be regulated, at least in some instances, by agonist occupancy of cell surface receptors. For example, in a smooth muscle cell line, either the activation of β-adrenergic receptors, or the addition of a cell-permeant analog of cAMP, results in a 70% reduction of (PP)2-IP4 concentration (but not that of either PP-IP4 or PP-IP5; see Safrany and Shears 1998). Atypically, this is not mediated via protein kinase A activation. A sustained depletion of intracellular Ca2+ in hepatocytes following the addition of thapsigargin has also been shown to reduce the steady state concentrations of diphosphoinositol polyphosphates by 50%. However, it seems unlikely that receptor activation of phospholipase C regulates the turnover of the diphosphoinositol polyphosphates, as the addition of vasopressin to hepatocytes does not alter the steady state concentrations of either PP-IP5 or (PP)2-IP4 (Glennon and Shears 1993).
There are three potential enzymatic routes for the dephosphorylation of the diphosphoinositol polyphosphates (Fig. 3). The first is the reversal of the IP6- and PP-IP5-kinase reactions, as outlined above. Secondly, the multiple inositol polyphosphate phosphatase can carry out the dephosphorylation (Craxton et al. 1997). This enzyme, first cloned from liver, is also enriched in brain. More recently, a group of enzymes known as the diphosphoinositol polyphosphate phosphatases (DIPP) have been identified. One of these is reported to be regulated by Li+ (Hua et al. 2001). The first member of this family to be isolated was an 18-kDa, Mg2+-dependent enzyme that exhibited high affinity for both PP-IP5 and (PP)2-IP4, and was present in brain along with other tissues (Safrany et al. 1998). It was subsequently demonstrated that PP-IP4 is also a substrate for DIPP (Caffrey et al. 2000). Two additional isoforms of DIPP (MW = 21 kDa) have been identified, DIPP2α and DIPP2β, the latter being distinguishable by the presence of an additional glutamine residue. This change results in an enzyme that is 65–80% less active against PP-IP5, PP-IP4 or (PP)2-IP4 than is the α-isoform (Caffrey et al. 2000). The purification and molecular cloning of DIPPs has revealed them to contain a Nudix (nucleoside diphosphate linked to some other moiety, X) motif. Nudix, also known as a MutT motif, is a 23 amino acid residue domain found in enzymes that protect cell integrity against oxygen radicals. The DIPPs are unusual in that they are the first example of a MutT-type enzyme that can readily metabolize a class of substrates other than nucleoside diphosphates. However, the diphosphoinositol polyphosphates do not appear to be general substrates for other MutT-type enzymes (see Safrany et al. 1999).
Although the first indication of the importance of inositol-containing phospholipids in cell signaling events was obtained nearly 50 years ago (Hokin and Hokin 1953, 1955), discovery of new roles for both the phosphoinositides and the water-soluble inositol phosphates continue unabated. In this review, we have focused on the importance of myo-inositol, the basic building block, and its water-soluble phosphates, in CNS function. Not only does inositol serve as precursor for the synthesis of inositol phospholipids, but it also plays an essential role in the regulation of cell volume within the CNS. The versatility of function of inositol and its phosphates is further illustrated by the case of IP6 and disphosphoinositol polyphosphates, highly phosphorylated forms of inositol that are readily synthesized in the brain and are now linked to both the regulation of nuclear function and to endo/exocytosis.
Inositol plays an essential role in eukaryotes, and the presence of inositol-containing phospholipids has been documented in some eubacteria and Archebacteria. Thus, it is likely that inositol was one of the earliest organic molecules to evolve, and that during the course of evolution it has been adapted to serve in many functions. Given the variety of possible inositol phosphates, it is likely that additional roles for inositol and its phosphates in the nervous system are yet to be uncovered.
The authors wish to thank Dion Frischer for her invaluable assistance in all aspects of the preparation of this manuscript. The work was supported in part by NIH grant NS 23831 (SKF), by the Colton Frontiers in Neuroscience fund (BWA) and by the Mental Health Research Institute of the University of Michigan.