The mood stabiliser lithium suppresses PIP₃ signalling in Dictyostelium and human cells

Bipolar mood disorder (manic depression) is a major, chronic psychiatric disorder with a lifetime prevalence of greater than 1%. Characterised by recurrent episodes of mania and depression, bipolar disorder presents a significant socio-economic burden (Das Gupta and Guest, 2002) and substantially lowers quality of life (Gutiérrez-Rojas et al., 2008). If untreated, it carries a high suicide rate, estimated to be 1 million deaths worldwide per year. The impact of bipolar disorder is not restricted to mental health, as it exhibits co-morbidity with cardiovascular, endocrine and metabolic disorders and is linked to an elevated inflammatory response. The molecular basis of the disorder is unknown. One of the few investigative routes available is to study the mechanism of action of mood stabiliser drugs. Lithium is the most established and commonly used mood stabiliser (Burgess et al., 2001); in addition, preliminary evidence indicates that lithium might suppress the effects of the neurodegenerative disease amyotrophic lateral sclerosis (ALS) (Fornai et al., 2008). In both cases, we lack therapeutic mechanisms. Establishing these mechanisms will lead to a better understanding of the molecular basis of bipolar mood disorder, a rationale for its use in the treatment of neurodegeneration and improved drug design; furthermore, the mechanisms may reveal other therapeutic uses for the existing mood stabilisers.

Inositol depletion is a common outcome of treatment with a number of mood stabiliser drugs (Berridge, 1985; Williams et al., 2002). Lithium inhibits the key inositol synthetic enzymes, inositol monophosphatase (IMPase) and inositol polyphosphate phosphatase (IPP) (Hallcher and Sherman, 1980), whereas an alternative mood stabiliser, valproic acid (VPA), inhibits inositol synthase activity (Shaltiel et al., 2004). A third mood stabiliser, carbamazepine, also causes inositol depletion, but by an unknown mechanism (Williams et al., 2002; Sarkar et al., 2005; Shimshoni et al., 2009). Both lithium and VPA lower the cellular concentration of the second messenger, inositol (1,4,5)-trisphosphate [IP₃, also known as Ins(1,4,5)P₃], leading to the proposition that IP₃ might be the important target in bipolar disorder. However, despite intensive investigation, it has not been possible to directly correlate altered IP₃ signalling and bipolar mood disorder. This raises the possibility that mood stabilisers may suppress alternative forms of inositide signalling.

The social amoeba Dictyostelium discoideum provides a good system in which to study the mechanism of action of mood stabiliser drugs and to test drug function. Dictyostelium development is lithium sensitive, but in contrast to other model systems, the two major lithium-sensitive signal pathways, inositol phosphate (IP) and glycogen synthase kinase-3 (GSK-3), are separated into different developmental stages. GSK-3 inhibition leads to mis-patterning of cell specification during multicellular stages of development, whereas inositol depletion substantially retards Dictyostelium cell aggregation (Maeda, 1970; Williams et al., 1999). Here, we demonstrate that phosphatidylinositol (3,4,5)-trisphosphate [PIP₃, also known as PtdIns(3,4,5)P₃] signalling is the major lithium target during Dictyostelium chemotaxis and is suppressed in the human neutrophil cell line HL60.

When starved, Dictyostelium amoebae enter a developmental programme to generate a multicellular fruiting body (Kessin, 2001). To achieve this, cells must aggregate to form a mound of 10⁵ cells; a process that requires cell chemotaxis towards cyclic adenosine 3′,5′-monophosphate (cAMP). Lithium treatment had no effect on cAMP synthesis (Fig. 1A), but did have a striking effect on chemotaxis (Fig. 1B; supplementary material Movies 1 and 2). Analysis of individual cells undergoing chemotaxis showed that lithium causes a decrease in cell speed, an increase in cell turning and a small decrease in their chemotactic index (CI), which is a measure of their ability to sense the signal gradient (supplementary material Table S1). A 56% mean decrease in cell speed correlates well with the doubling of aggregation time that has been observed previously for lithium-treated cells (Williams et al., 1999). In the majority of cases, lithium was applied acutely for 1 hour as 10 mM LiCl, however the same effect could be achieved after prolonged treatment in 2 mM LiCl (data not shown). Dictyostelium cells move by local assembly of the F-actin cytoskeleton at their leading edge. Lithium did not interfere with the basic mechanisms of F-actin polymerisation (Fig. 1C), suggesting that the lithium targets an intracellular signalling mechanism.

The effects of lithium treatment on chemotaxis strongly resembled those reported for reduced PIP₃ signalling (Loovers et al., 2006; Takeda et al., 2007). PIP₃ is rapidly induced by cAMP at the leading edge of the cell through translocation of class I phosphoinositide 3-kinases (PI3K) (Funamoto et al., 2002). Although not essential, PIP₃ is required for efficient chemotaxis (Andrew and Insall, 2007; Hoeller and Kay, 2007). Under the conditions used here, we found that treatment with the PI3 kinase inhibitor LY294002 closely matches that seen with lithium (Fig. 1B). A similar defect was seen for mutant cells lacking all five Dictyostelium PI3K genes (pi3k1-5, also known as pikA-C, pikF and pikG) (Hoeller and Kay, 2007) (Fig. 1B).

To probe the effect of lithium on PIP₃ signalling, we monitored PIP₃ synthesis in living cells by expression of the PIP₃-specific binding protein PH_CRAC-GFP (Parent et al., 1998). When globally stimulated, PIP₃ is generated on the entire plasma membrane causing translocation of PH_CRAC-GFP. Lithium treatment suppressed PH_CRAC-GFP translocation to 40% of that seen in control cells (Fig. 2A). To confirm these observations, we monitored phosphorylation of the protein kinase B (PKB) homologue, PkbA, an event that is dependent on its membrane translocation through PH domain binding to PIP₃ (Kamimura et al., 2008). In control cells, phospho-PkbA (p-PkbA) could be detected 10 seconds after cAMP stimulation and was lost after 45 seconds (Fig. 2B,C). PkbA phosphorylation was not observed in the pi3k1-5-null mutant or in the pkbA-null mutant (Fig. 2B). Lithium treatment suppressed PkbA phosphorylation to 45% of control values (Fig. 2C). Finally, we directly measured PIP₃ synthesis following cAMP stimulation and found that new PIP₃ production was suppressed to 38% of that seen following control treatment; this is consistent with the observations on PkbA phosphorylation (Fig. 2C). Together, these results demonstrate that lithium suppresses PIP₃ signalling following cAMP stimulation and can account for the effect of lithium on chemotaxis.

Although PIP₃ signalling was suppressed following cAMP stimulation, other intracellular signalling pathways that are required during chemotaxis were unaffected. Dictyostelium contains two homologues of the protein kinase PKB/AKT1, PkbA and PkgB (also known as PKBR1) (Meili et al., 2000). Both are phosphorylated in response to cAMP stimulation, however only PkbA phosphorylation is dependent on PIP₃ (Kamimura et al., 2008). Consistent with a specific effect on PIP₃, we found that PkgB phosphorylation is unaffected by lithium treatment (Fig. 2B). Cyclic guanosine 3′,5′-monophosphate (cGMP) mediates an additional independent regulator of chemotaxis (Veltman and van Haastert, 2008), however lithium has no effect on cGMP synthesis (Fig. 2D). Finally, ablation of the Dictyostelium PLC (plc) and IP₃ receptor (iplA) genes does not suppress chemotaxis (Drayer et al., 1994; Traynor et al., 2000) (Fig. 1B and data not shown), indicating that inhibition of IP₃ production or calcium release by lithium is not related to the effects of lithium on chemotaxis.

Following cAMP stimulation, PI3K translocates to the leading edge of the cell, whereas the PIP₃ phosphatase PTEN is lost from this region. Lithium had no effect on either PI3K translocation to the plasma membrane or PTEN loss from the membrane following cAMP stimulation (Fig. 3E; supplementary material Fig. S1A), indicating that cAMP-stimulated PIP₃ synthesis is not suppressed by loss of PI3K or elevated PTEN activity. We therefore investigated a third possibility, which was whether lithium reduced the amount of the PI3K substrate phosphatidylinositol (4,5)bisphosphate [PIP₂, also known as PtdIns(4,5)P₂]. We could not detect any lithium-induced changes in the steady-state concentration of PIP₂, however we did detect a 50% and 63% decrease in the rate of PIP and PIP₂ synthesis, respectively (Fig. 3A,B; supplementary material Fig. S1B); this level of reduction closely matches that calculated for cAMP-induced PIP₃ synthesis. These observations suggest that there are two pools of PIP₂. First, there is a slowly metabolised pool that is not involved in cell signalling during chemotaxis and that is insensitive to lithium treatment. Most of this PIP₂ is likely to be bound to proteins, particularly those regulating the actin cytoskeleton (Janmey and Lindberg, 2004). Second, there is a rapidly metabolised pool, which is required for the synthesis of PIP₃ during chemotaxis and is susceptible to lithium inhibition. In Dictyostelium, the amount of PIP₂ within the lithium-insensitive pool is much greater than that involved in cell signalling; we calculated that there was a 50- and 200-fold excess of total PIP₂ concentration over IP₃ and PIP₃, respectively. The presence of a large non-dynamic PIP₂ pool that masks the effects of lithium on PIP₂-mediated signalling might be one reason why, in many systems, it has proven difficult to measure substantial changes in PIP and PIP₂.

The behaviour of Dictyostelium PTEN offers a means to directly monitor the rapidly metabolised, lithium-sensitive PIP₂ pool. In Dictyostelium, PTEN localisation to the plasma membrane is dependent on PIP₂, and a reduction in PIP₂ leads to a reduced rate of PTEN binding (Vazquez et al., 2006). Lithium treatment of unstimulated cells reduced the amount of PTEN-GFP associated with the plasma membrane (Fig. 3C), consistent with a reduction of PIP₂. In agreement with these observations, lithium treatment of unstimulated cells increased PIP₃ and p-PkbA compared with control-treated cells (Fig. 3D).

Although a good marker for changes in the PIP₂ signalling pool, the change in PTEN behaviour following lithium treatment is unlikely to have a major effect on cell signalling. This is because cAMP stimulation removes PTEN from the plasma membrane and hence stimulated production of PIP₃ arises solely from PI3K activation. As lithium treatment only alters the amount of bound PTEN and not its relocation from the membrane (Fig. 3E), the reduced amount of newly synthesised PIP₃ following cAMP stimulation is probably the result of suppressed PIP₃ generation. During chemotaxis, cells generate a steep gradient of PIP₃, with high concentrations at the leading edge of the cell, and lithium treatment will act to flatten this gradient. The effects we observe following lithium treatment match the phenotype of pi3k1-5-null mutants and not that of pten-null mutants (Iijima and Devreotes, 2002; Hoeller and Kay, 2007; Kortholt et al., 2007), suggesting that a failure of PIP₃ synthesis, rather than reduced breakdown of the protein, gives rise to the effect of lithium.

A number of experiments suggest that the effects of lithium on phosphoinositide signalling may have arisen through inositol depletion. First, propylisopropylacetic acid (PIA), a specific inhibitor of myo-inositol synthesis (Shimshoni et al., 2007), has a similar effect on chemotaxis to lithium (Fig. 1B). Second, IMPase is inhibited by lithium through an uncompetitive mechanism (Atack et al., 1995), which means that the inhibition can be reversed by increasing the concentration of the enzyme. Consistent with such a mechanism, lithium sensitivity is suppressed by overexpression of the Dictyostelium IMPase gene (impa1) (Fig. 1B). This lithium resistance can also be observed at the molecular level where impa1 overexpression counteracts the effect of lithium on PkbA phosphorylation and reverses the inhibition of PIP/PIP₂ synthesis (Fig. 3B,F).

These observations indicate that PIP₃ is the major target of lithium in Dictyostelium chemotaxis. To examine whether this is unique to Dictyostelium, we investigated the effect of lithium on the human cell line HL60 following their differentiation into neutrophil-like cells that are capable of chemotaxis (Collins et al., 1979). These differentiated HL60 cells undergo chemotaxis to the peptide fMLP in a process that involves PIP₃ (Cui et al., 2000). Global stimulation using a chemoattractant generates PIP₃ on the plasma membrane, which we monitored by plasma membrane translocation of the PIP₃-binding protein PH_Akt-GFP (Servant et al., 2000) (Fig. 4A). Consistent with our Dictyostelium observations, PH_Akt-GFP translocation was attenuated following lithium treatment (Fig. 4B). This indicates that the effects of lithium on PIP₃ signalling are conserved in human cells, extending the repertoire of the effects of lithium on inositide signalling.

In conclusion, we have shown that lithium suppresses PIP₃ signalling. These results have implications beyond the regulation of chemotaxis and could explain the therapeutic efficacy of lithium in the treatment of bipolar disorder. PIP₃ signalling in the nervous system is important for axonal guidance, synaptogenesis and synaptic transmission (Zhou et al., 2004; Ramsey et al., 2005; Chadborn et al., 2006; Martin-Pena et al., 2006; Xu et al., 2007), and mice with elevated PIP₃ levels suffer seizures (Backman et al., 2001). Lithium might, therefore, alter these processes through an effect on PIP₃ signalling. Furthermore, bipolar mood disorder has a strong genetic predisposition but is a complex genetic trait that probably involves combinations of multiple loci (Craddock and Jones, 1999). As yet, no susceptibility gene has been unambiguously identified but, interestingly, many of the candidate genes currently under investigation converge on the PIP₃- and PKB-mediated signal pathway (Silverstone et al., 2005; Carter, 2007). The results presented here provide a mechanism that might directly couple current genetic studies with a target of lithium therapy.

Conversely, suppression of PIP₃ signalling might open up therapeutic uses for lithium. In human patients, aberrant PIP₃ signalling is linked to macrocephaly, mental retardation, cerebellar hypertrophy, ataxia, seizures, epilepsy and autism (Li et al., 1997; Yates, 2006). Beyond the nervous system, regulators of PIP₃ signalling are frequently mutated in tumours (Leslie and Downes, 2004; Barber and Welch, 2006) and familial PTEN mutations are associated with a number of syndromes that cause hamartomas, such as Cowden, Bannayan-Riley-Ruvalcaba and Proteus syndromes (Eng, 2003). The discovery that suppression of PIP₃ signalling is a major consequence of lithium treatment offers both an insight into the mechanism of lithium action within the nervous system and suggests therapeutic opportunities for this well-established drug, such as in the control of tumour progression.

D. discoideum strain AX2 was used in all assays. Washed log-phase cells were developed in shaking suspension in KK2 buffer for 5 hours at 5×10⁶ cells/ml with 6-minute pulses of 100nM cAMP. All drug and control treatments were added only for the final hour, except for PIA, which is required throughout pulsing. Pulsed cells were placed in a Z02 Zigmond chamber (Neuro Probe) (Zigmond, 1977) with 1 μM of cAMP/KK2 as the source and KK2 as the sink. All solutions contained the appropriate drugs to ensure continued treatment. Cells were recorded after 20 minutes using video microscopy with a 20× objective and by taking DIC images every 6 seconds. Cell movement was analysed using the Dynamic Image Analysis System (DIAS) (version 3.4.1, Soll Technologies, Iowa City, IA) (Soll et al., 2001). For statistical analysis, we used Kruskal-Wallis and Mann-Whitney U tests, with a post-hoc Dunn test for multiple comparison.

cAMP and cGMP were measured as described previously (Snaar-Jagalska and Van Haastert, 1994) using isotope dilution assay kits (GE healthcare). Actin polymerisation was measured using a modified version of the method of Hall et al. (Hall et al., 1988). Pulsed cells (2.5×10⁷ cells/ml) were stimulated with 1 μM cAMP and 100 μl aliquots were fixed at the set time intervals by addition of 900 μl of stop solution (3.7% formaldehyde, 0.1% Triton X-100, 0.5 μM TRITC-phalloidin, 10 mM PIPES, 5 mM EGTA, 2 mM MgCl₂, pH 6.8). Cells were resuspended in 800 μl of methanol and F-actin was assayed as fluorescence intensity at 540/590 nm. Wild-type cells were transformed by electroporation with the plasmids WF38 (PH_CRAC-GFP) (Parent et al., 1998), PTEN-GFP (Iijima and Devreotes, 2002), pOH30 (YFP-PI3K2N) or pPH_PLC (PLC-δ1-PH-GFP) (Hoeller and Kay, 2007). Pulsed cells were stimulated by the addition of 1 μM of cAMP. Protein translocation was recorded by fluorescence video microscope with a 60× objective, and the fluorescence intensity of the cytosol of each cell in each frame was measured using ImageJ software and normalised to the prestimulated level of the same cell. For PKB phosphorylation, pulsed cells were stimulated with 1 μM of cAMP and samples were lysed directly into 100 μl of 2×LDS loading buffer (Invitrogen) supplemented with 100 mM NaF, 0.4 mM Na₃VO₄ and 3 mM EDTA on ice. Samples were separated by electrophoresis using 3–8% NuPage tris-acetate gels (Invitrogen), blotted onto nitrocellulose membrane (Amersham) and then probed with a rabbit anti-phospho PKC (pan) antibody (190D10, Cell Signalling Technology) or, for quantification, with an anti-phosphothreonine antibody (Lim et al., 2001). PIP₃ was measured directly by PIP₃ mass ELISA (Echelon). Phosphoinositide turnover was monitored by incubating chemotactic-competent cells in internal buffer [139 mM sodium glutamate, 5 mM glucose, 5 mM EGTA, 20 mM PIPES at pH 6.6, 1 mM magnesium sulphate, 0.03% saponin, 0.005 volume of protease inhibitor cocktail 1 and 2 (Sigma), and 5 μCi of γ-³²P-labelled adenosine triphosphate (ATP)] for exactly 6 minutes, followed by lysis in acidified methanol, extraction in water-methanol-chloroform (1:2:1), and separation by thin-layer chromatography on silica gel 60 TLC plates (VWR) in a solvent mix (chloroform-acetone-methanol-acetic acid-distilled water in a ratio of 40:15:13:12:7). ³²P lipid incorporation into lipids was measured using a Typhoon phosphoimager (Pawolleck and Williams, 2009).

HL60 cells were grown at 37°C in RPMI 1640 medium supplemented with 25 mM HEPES, 10% fetal bovine serum, 100 μg/ml streptomycin, 100 U/ml penicillin and 2 mM L-glutamine. Before stimulation, the cells were differentiated into neutrophil-like cells in 8-well chamber slides (Lab-Tek) by the addition of 1.3% DMSO to the medium for 5 days. Cells were stimulated by pipetting fMLP (Sigma) directly into each chamber, to a final concentration of 1 μM. Translocation of the GFP marker to the membrane was quantified as for Dictyostelium cells, with images taken at 5 second intervals.

This work was supported by a Wellcome Trust Programme Grant to A.J.H. We wish to thank Maurice Hallet and Trevor Dale for helpful discussions.