Lithium and GADL1 regulate glycogen synthase kinase-3 activity to modulate KCTD12 expression

Potassium channel tetramerization domain containing 12 (KCTD12), the auxiliary GABAB receptor subunit, is identified as a susceptibility gene for bipolar I (BPI) disorder in the Han Chinese population. Moreover, the single-nucleotide polymorphism (SNP) rs17026688 in glutamate decarboxylase–like protein 1 (GADL1) is shown to be associated with lithium response in Han Chinese BPI patients. In this study, we demonstrated for the first time the relationship among lithium, GADL1, and KCTD12. In circulating CD11b+ macrophage cells, BPI patients showed a significantly higher percentage of KCTD12 expression than healthy controls. Among BPI patients, carriers of the ‘T’ allele (i.e., CT or TT) at site rs17026688 were found to secrete lower amounts of GADL1 but higher amounts of GABA b receptor 2 (GABBR2) in the plasma. In human SH-SY5Y neuroblastoma cells, lithium treatment increased the percentage of KCTD12 expression. Through inhibition of glycogen synthase kinase-3 (GSK-3), lithium induced cyclic AMP-response element binding protein (CREB)–mediated KCTD12 promoter activation. On the other hand, GADL1 overexpression enhanced GSK-3 activation and inhibited KCTD12 expression. We found that lithium induced, whereas GADL1 inhibited, KCTD12 expression. These findings suggested that KCTD12 may be an important gene with respect to neuron excitability and lithium response in BPI patients. Therefore, targeting GSK-3 activity and/or KCTD12 expression may constitute a possible therapeutic strategy for treating patients with BPI disorder.

For bipolar patients, lithium is the first-line choice for maintenance treatment since it can reduce the risk of relapse and suicide [1][2][3] . However, only 30% of patients have an excellent response to lithium with complete remission of symptoms, as has been observed for patients of European descent 4,5 . Glutamate decarboxylase-like protein 1 (GADL1) has aspartate 1-decarboxylase and cysteine sulfinic acid decarboxylase activities to produce β-alanine, hypotaurine, and taurine 6 . Chronic administration of lithium is found to decrease the level of taurine in the rat brain 7,8 , and the enzyme activity of GADL1 increases significantly in the presence of lithium 9 . The single-nucleotide polymorphism (SNP) rs17026688 in GADL1 has been found to be associated with lithium response in bipolar I (BPI) patients of Han Chinese descent. Patients carrying the allele 'T' (i.e., CT or TT) at rs17026688 are lithium good responders, while those carrying the homozygous allele C are lithium poor responders 10 . The SNP rs17026688 T carriers have lower frequencies of recurrent episodes than non-T carriers when these patients are compared during the cumulative period of good drug adherence 11 . However, the association between rs17026688 and lithium response has not been replicated with other clinical samples from different human populations 12,13 . Therefore, the role of GADL1 in the neuropsychiatric diseases and lithium response requires further investigation.
The gene KCTD12, encoding potassium channel tetramerization domain containing 12, is highly associated with BPI disorder in the Han Chinese population 14 . KCTD12, one of the auxiliary GABA B receptor subunits, can increase the GABA B receptor expression on the cell surface and the magnitude of downstream signaling 15,16 .
GABA B receptors, G-protein-coupled receptors for GABA, regulate neuronal excitability in the mammalian nervous system. Thus, GABA B receptors are involved in neurological and psychiatric diseases, including epilepsy, schizophrenia, depression, and anxiety 17,18 . Notably, GABA B receptors, but not GABA A receptors, are upregulated in the hippocampus and frontal cortex after chronic lithium treatment in rats 19,20 .
The level of GABA, the main inhibitory neurotransmitter in the central nervous system, has been reported to be low in the plasma and cerebrospinal fluid of patients with mood disorders [21][22][23][24] . In euthymic bipolar patients, the use of lithium as a mood stabilizer is found to increase the level of GABA in both plasma and cerebrospinal fluid 21,25,26 . We hypothesized that lithium or GADL1 could regulate KCTD12 expression, and herein investigated the mechanism underlying the regulation of KCTD12 expression by lithium and GADL1 in the human neuroblastoma cells, SH-SY5Y.
The activity of glycogen synthase kinase-3 (GSK-3) is regulated by phosphorylation. For example, phosphorylation at Ser9 (pSer9) of GSK-3β results in its inactivation, whereas pTyr279 of GSK-3α or pTyr216 of GSK-3β results in the activation of these GSKs 27 . Lithium can inhibit the activity of GSK-3, leading to release of several transcription factors into the nucleus, including cAMP response element binding protein (CREB), heat-shock factor-1, and β-catenin 28 . In this study, we addressed how lithium and GADL1 influenced the activity of GSK-3, which regulated the expression of KCTD12 using the GADL1 stable overexpression neuroblastoma cell line.

Results
The expression of GADL1, taurine, GABA, GABA B receptor 2 (GABBR2) and KCTD12 among BPI patients and healthy controls. First, we compared plasma levels of GADL1 and its catalytic product, taurine, in BPI patients and healthy controls, showing that BPI patients secreted significantly higher amounts of GADL1 than healthy controls in the plasma (Table 1). Next, we compared their secretions between T and non-T carriers at rs17026688 among BPI patients or healthy controls. In both BPI patients and healthy controls, non-T carriers had significantly higher levels of GADL1 ( Supplementary Fig. S1a) and taurine ( Supplementary Fig. S1b) than T carriers (Table 1).
KCTD8, KCTD12, and KCTD16 bind to GABBR2 as part of a stable receptor complex 15,16 . We measured the plasma levels of GABA and GABBR2 among BPI patients and healthy controls. BPI patients secreted significantly higher amounts of GABBR2 than healthy controls in the plasma (Table 1). Next, we compared GABA and GABBR2 secretions between T and non-T carriers at rs17026688 among BPI patients or healthy controls. T carriers secreted higher amounts of GABA than non-T carriers among healthy controls although we found no significant difference between T and non-T carriers among BPI patients (Supplementary Fig. S1c and Table 1). T carriers (118.73 ± 35.04 pg/ml) of healthy controls also secreted significantly higher levels of GABA than T carriers (103.90 ± 31.92 pg/ml) of BPI patients, as analyzed by Mann-Whitney test with p = 0.0238 ( Supplementary  Fig. S1c). On the other hand, T carriers secreted significantly higher amounts of GABBR2 than non-T carriers among BPI patients, whereas no significant difference was found between T and non-T carriers among healthy controls (Supplementary Fig. S1d and Table 1). T and non-T carriers of BPI patients secreted significantly higher amounts of GABBR2 than those of healthy controls, as analyzed by Mann-Whitney tests with p = 0.0002 and p = 0.0006, respectively ( Supplementary Fig. S1d).
Microglia cells, the glia cells and macrophages in the brain, can mediate neuroinflammation and express many types of neurotransmitter receptors, including GABA B receptors 29,30 . Thus, we further characterized CD11b + macrophage cells with respect to their KCTD12 expression. BPI patients showed a significantly higher percentage of KCTD12 expression in macrophage cells than healthy controls (Table 1). In the gated CD11b + macrophage cells, the percentage of KCTD12 expression did not differ between rs17026688 T and non-T carriers among BPI patients or healthy controls (Table 1), though T carriers (13.74 ± 13.57%) of BPI patients showed a significantly higher percentage of KCTD12 expression than T carriers (7.62 ± 4.66%) of healthy controls, as analyzed by Mann-Whitney test with p = 0.0047 ( Supplementary Fig. S1e). In comparison, no significant difference was found between non-T carriers of BPI patients and those of healthy controls ( Supplementary Fig. S1e). Taken together with Supplementary Fig. S1d, rs17026688 T carriers of BPI patients showed a more neuro-inhibitory status, as reflected by the plasma levels of GABBR2.  Table 1. Plasma or PBMC detection in BPI patients and healthy controls. Data are shown as mean (%) ± S.D. The percentage of KCTD12 expression was analyzed in the gated macrophage cells. The differences between healthy controls and BPI patients or between T and non-T carriers among BPI patients or healthy controls were calculated by 1-tailed Mann-Whitney test (*p < 0.05; **p < 0.01; ***p < 0.001).
Identification of cAMP-responsive elements (CREs) in the KCTD12 promoter. We further explored the mechanisms for our findings that lithium increased KCTD12 expression in SH-SY5Y cells. CREB mediates the activation of cAMP-responsive genes by binding to one of the conserved CREs, TGACGTCAA, TGACG (half-site), or TGANNT(CA) 31,32 . Analysis of the KCTD12 promoter revealed two CREs in the 869 bp upstream of the transcription start site ( Supplementary Fig. S2a), suggesting that lithium-induced upregulation of KCTD12 expression is likely mediated through CREB binding to KCTD12 promoter. In fact, chromatin immunoprecipitation (ChIP) and luciferase reporter assays revealed that CREB could bind to the KCTD12 promoter ( Supplementary Fig. S2b) and thereby activate KCTD12 promoter-driven luciferase activity ( Supplementary  Fig. S2c).

Inhibition of GSK-3 by lithium results in the upregulation of KCTD12 transcription. A luciferase
reporter assay was used for elucidating the signaling pathways underlying the lithium-induced, CREB-mediated upregulation of KCTD12 (Fig. 2). SH-SY5Y cells were co-transfected with the CREB-EGFP plasmid and a luciferase reporter linking to KCTD12 promoter. At 7 h prior to harvest, cells were treated with LiCl or the GSK-3β inhibitor, SB415286, in the indicated groups (Fig. 2a). Either lithium (the 2 nd group) or SB415286 (the 3 rd group) could significantly upregulate KCTD12 promoter-driven luciferase transcription. To test whether lithium can influence cAMP-induced transcription, SH-SY5Y cells were treated with 8-bromoadenosine cAMP (8brcAMP), an analog of cAMP that has greater stability and increased membrane permeability (Fig. 2a). 8brcAMP alone (the 4 th group) increased CREB-mediated KCTD12 transcription to a substantial degree (p = 0.02, as analyzed by the student t test). 8brcAMP also significantly inhibited the lithium-induced increase in the KCTD12 transcription (the 5 th group), suggesting that there were no synergistic effects in the presence of lithium and 8brcAMP.
In addition to GSK-3 33 , lithium salts are known to inhibit inositol monophosphatase (IMP) and thus deplete inositol in cells 34 . Therefore, the lithium-treated cells were co-treated with inositol to replenish the presumed depleted stores of inositol (the 7 th group in Fig. 2a). The addition of inositol did not reverse the effect of lithium on KCTD12-driven luciferase activity (the 7 th group). Inositol alone (the 6 th group) could induce CREB-mediated KCTD12 transcription to a substantial degree (p = 0.007, as analyzed by the student t test). Taken together with Supplementary Fig. S2, these data indicated that lithium-induced, CREB-mediated KCTD12 transcription acts through GSK-3β inhibition, but not through activation of cAMP-protein kinase A (PKA) pathway or suppression of IMP activity, as shown in Fig. 2b. www.nature.com/scientificreports www.nature.com/scientificreports/
Downregulation of KCTD12 mRNA level in the GADL1-overexpressing cells. Total RNA extracted from SH-SY5Y and GADL1-overexpressing cells was analyzed with an RNA expression array, revealing that GADL1 was overexpressed in the stable clone as compared with the parental SH-SY5Y cell line; however, KCTD12, KCTD16, and CREB5 were downregulated (Fig. 4a). Real-time quantitative PCR (RT-qPCR) was then performed to validate the RNA expression array data. Indeed, GADL1 was upregulated (2.48-fold increase), whereas KCTD12, KCTD16, and CREB5 were downregulated compared with SH-SY5Y cells (Fig. 4b).
These data together with Supplementary Fig. S1 and Table 1, suggest a model for lithium nonresponsiveness in non-T carriers among BPI patients (Fig. 4g). Lithium upregulates KCTD12 expression and then promotes expression of GABA B receptor, which strengthens downstream G-protein-coupled (G βγ ) signaling. In contrast, GADL1 overexpression inhibits KCTD12 expression. Non-T carriers express higher amounts of GADL1 but lower amounts of KCTD12, probably leading to increased neuron excitability and contributing to the nonresponsiveness to lithium treatment.

Discussion
In this study, we found that BPI patients expressed a higher percentage of KCTD12 expression in macrophage cells than healthy controls and that rs17026688 T carriers secreted lower amounts of GADL1 and taurine than non-T carriers among Han Chinese BPI patients. Furthermore, we addressed for the first time the effects of lithium and GADL1 on the regulation of KCTD12 expression in human neuroblastoma cells. GADL1 catalyzes the www.nature.com/scientificreports www.nature.com/scientificreports/ decarboxylation of aspartate, cysteine sulfinic acid, and cysteic acid to produce β-alanine, hypotaurine, and taurine 6 . In rats, chronic administration of lithium decreases the level of taurine in the brain 7,8 . This observation in rats may provide a hint for our findings that, among Han Chinese BPI patients, rs17026688 T carriers had lower amounts of secreted GADL1 and taurine than non-T carriers. Besides, we found that BPI patients secreted higher amounts of GADL1 than healthy controls in the plasma, suggesting that GADL1 might play a role in the development of bipolar disorder in the Han Chinese population. However, we acknowledged the limitation that possible confounding factors were not controlled due to the exploratory nature of this study and small sample sizes.
Monocytes can transform into microglia cells when circulating to the brain 35 . Microglia cells, the glia cells and macrophage in the brain, can mediate neuroinflammation and bear many types of neurotransmitter receptors including GABA B receptors on their cell surface 29,30 . In fact, KCTD12 is highly expressed in mouse brain microglial cells 36 . In the brain, microglia cells have effects on bipolar disorder during disease development 37 . We found that BPI patients expressed higher levels of GABBR2 in the plasma and a higher percentage of KCTD12 expression in macrophage cells than healthy controls. These observations suggested that KCTD12 and/or GABA signaling pathway might be involved in the disease progression of bipolar disorder, which echoed the previous finding that KCTD12 is a risk gene for BPI disorder in the Han Chinese population 14 .  www.nature.com/scientificreports www.nature.com/scientificreports/ In addition to KCTD12, T carriers secreted greater amounts of GABBR2 than non-T carriers although the plasma levels of GABA did not differ significantly between T and non-T carriers among BPI patients. Treatment with lithium has been reported to trigger an increase or no changes in the plasma levels of GABA in bipolar patients, and the amounts of rat brain GABA B receptors may be increased or decreased after lithium treatment 26,38 . Among healthy controls, rs17026688 T and non-T carriers showed significant differences on the secretion of GADL1, taurine, and GABA in the plasma in this study, suggesting that the SNP rs17026688 itself had influence on the plasma levels of GADL1, taurine, and GABA secretions, which might not be related with bipolar disorder or lithium drug use.
Besides peripheral blood cells, we examined human neuroblastoma cells since most neurons in the brain express GABA B receptors and at least one KCTD protein 15,39 . Moreover, GADL1 expression amounts are more in neurons than in glia cells in the human adult brain 9 . Indeed, lithium increased the percentage of KCTD12 expression in SH-SY5Y cells in our study. We hypothesized that through the upregulation of KCTD12 expression in neurons, lithium might strengthen GABA B receptor signaling and reduce the neuronal excitotoxicity in the brain so as to maintain mood stability. However, this hypothesis on lithium action needs validation in the future.
We further elucidated the role of lithium in the induction of KCTD12 expression in SH-SY5Y cells. Lithium can inhibit the activity of GSK-3, leading to release of several transcription factors, including CREB, heat-shock factor-1, and β-catenin 28 . Analysis of the KCTD12 promoter revealed two CREs in the 869 bp upstream of the transcription start site. KCTD12 is the target gene of replication and transcription activator, a transcription activator of the gamma-herpesvirus family. This activator can form a complex with CREB, thereby activating or inhibiting CREB-response genes depending on the promoter context 40 . This evidence indirectly demonstrates that KCTD12 is a CREB-responsive gene. Indeed, CREB could bind to the KCTD12 promoter in both ChIP and luciferase assays. Further analysis of downstream signaling events revealed that, lithium-induced, CREB-mediated KCTD12 transcription acts through GSK-3 inhibition.
GSK-3 contains two isoforms, alpha and beta, both of which are inhibited by lithium 41 . Our flow cytometry analysis also showed that lithium could inhibit the phosphorylation of Tyr279 of GSK-3α and/or Tyr216 of GSK-3β in SH-SY5Y and GADL1 overexpression cells. SNPs in GSK-3β have been reported to be associated with lithium response 42 and the age at onset 43 in bipolar patients. GADL1 overexpression promoted GSK-3 activation and inhibited KCTD12 expression in our study. These findings suggested that targeting GSK-3 and/or KCTD12 expression may constitute a possible therapeutic strategy for treating patients with BPI disorder. Actually, many mood stabilizers (e.g. valproate) and anti-psychotic drugs for treating bipolar patients have impacts on GSK-3 and related signaling events 44 .
KCTD12 was originally identified as a susceptibility gene for BPI disorder in the Han Chinese population 14 . Here, we found that BPI patients expressed a higher percentage of KCTD12 expression in macrophage cells than healthy controls. In an earlier study, Kctd12 knockout mice show altered emotionality, behavior, and neuronal excitability 45 . Our present study also demonstrated for the first time the relationships among lithium, GADL1, and KCTD12 in human neuroblastoma cells. Lithium increased the percentage of KCTD12 expression in SH-SY5Y cells. The effects of lithium on the induction of KCTD12 expression were mediated though inhibition of GSK-3. Lithium-induced KCTD12 promoter activation may contribute to the molecular mechanism underlying its therapeutic effects in the T carriers of BPI patients. In comparison, GADL1 overexpression enhanced GSK-3 activation and inhibited KCTD12 expression, which might weaken downstream G βγ signaling. Non-T carriers expressed higher amounts of GADL1 but lower amounts of KCTD12, probably leading to more excitability in neurons and contributing to the observed lithium nonresponsiveness in these patients (Fig. 4g).

Study subjects. For immune endophenotype analysis, 76 BPI patients in remission (38T carriers and 38
non-T carriers) were recruited from the psychiatric departments of general hospitals and psychiatric institutions in Taiwan. A total of 60 healthy controls (31T carriers and 29 non-T carriers) were also recruited for comparisons. Their demographic characteristics are shown in Supplementary Tables S1. BPI disorder was diagnosed according to guidelines of the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (known as DSM-IV). Patients with other psychoses or affective disorders were excluded.
The procedures to recruit bipolar patients for this study were the same as previously described 10 . In short, psychiatric nurses and psychiatrists evaluated the study patients using a cross-culturally validated Chinese version of the Schedules for Clinical Assessment in Neuropsychiatry (known as SCAN) 46 and supplemented with available medical records and reports from family members and psychiatrists. All patients were euthymic at the time of blood collection. This study was approved by the institutional review board at each participating hospital and at Academia Sinica, Taiwan.
Ethics approval for this study was approved by the ethical committee of Chang Gung Medical Foundation, Mackay Memorial Hospital, Yuli Hospital, Ministry of Health and Welfare, Tsao-Tun Psychiatric Center, Ministry of Health and Welfare, Bali Psychiatric Center, Ministry of Health and Welfare, China Medical University and Hospital, and Academias Sinica, Taiwan. Informed consents were signed by enrolled patients and healthy controls. All experiments were performed in accordance with relevant guidelines and regulations. contrast, GADL1 inhibits KCTD12 expression and weakens downstream G βγ signaling. Non-T carriers express higher amounts of GADL1 and lower amounts of KCTD12, thereby downregulating inhibitory G βγ signaling, probably contributing to the observed lithium nonresponsiveness in these patients.
www.nature.com/scientificreports www.nature.com/scientificreports/ Genotyping. Genomic DNA of blood samples was purified using Genomic DNA Purification kit (Qiagen, USA). Amplification-refractory mutation system (ARMS) PCR was used to tell the genotypes at rs17026688 in the beginning. The inner primers used to tell the polymorphisms at rs17026688 were 5′-CATAAAATAAT TAGCATGCAAACAT TGGATAT T TC-3′ (for ward) and 5′-CCTGTCCTCACTAATGTATGAAGATCA-3′ (reverse), giving the band products of 174 and 286 bp for C and T, respectively. The outer primers used to check the success of PCR reaction were 5′-GATCAGACACTTGACCAATCTTGTTTAA-3′ (forward) and 5′-TTTGAGGGAATATATCAAGTGAAGTGTG -3′ (reverse), giving the band product of 432 bp. Direct sequencing and TaqMan SNP probe (C__34355332_10, Thermo Fisher) were performed to further validate the genotypes at rs17026688 as described elsewhere 10 .
Luciferase reporter assay. Using the jetPRIME transfection reagent (Polyplus), SH-SY5Y cells were transfected with a Renilla luciferase reporter plasmid carrying 869 bp of the KCTD12 promoter (SwitchGear), a firefly luciferase reporter plasmid, and a plasmid encoding CREB1-GFP (CREB, cyclic AMP-responsive element binding protein; GFP, green fluorescent protein) or control plasmid (pEGFP-C1, Clontech). After serum starvation overnight, a specific drug or inhibitor (LiCl, myoinositol, 8-bromoadenosine cAMP (8brcAMP) all from Sigma; SB415286 from Selleckchem) was added for different periods of time. Cells were lysed in reporter lysis buffer (Promega) containing a protease inhibitor cocktail (complete, EDTA-free, Roche) 2 days after transfection. Cell lysates were prepared to measure the luminescence of Renilla luciferase and firefly luciferase using a GloMax microplate scintillation and luminescence counter (Promega). The Renilla luciferase-derived luminescence from the KCTD12 promoter or control (Prom vector) was normalized to the luminescence measured from firefly luciferase, which accounted for differences in the transfection efficiency. siRNA knockdown in the GADL1-overexpressing cell line. GADL1-overexpressing cells were transfected with RISC-free negative control siRNA or siRNA targeting GADL1 at 0.1 μM using DharmaFECT1 transfection reagent 24 hr after cell seeding, as previously described 47 . Medium was changed 24 hr after transfection. Two days post transfection, cells from sextuplicate wells were harvested and pooled for subsequent RNA extraction and reverse transcription, followed by RT-qPCR analysis for GADL1, KCTD12, KCTD16, and CREB5. The fold-change value for each gene was normalized to ACTB expression. These assays were done in two independent experiments. Statistical analysis. Statistical differences between healthy controls and BPI patients or between T and non-T carriers among BPI patients or healthy controls were calculated by Mann-Whitney tests. All statistical tests were considered significant at p < 0.05 level. GraphPad Prism 5 software was used to draw the data distribution in the figures.

Data Availability
The RNA expression array datasets generated and analyzed in this study are available from the corresponding authors on reasonable request.