Human myo-inositol monophosphatase 2 rescues the nematode thermotaxis mutant ttx-7 more efficiently than IMPA1: functional and evolutionary considerations of the two mammalian myo-inositol monophosphatase genes


Address correspondence and reprint requests to Tetsuo Ohnishi, Laboratory for Molecular Psychiatry, RIKEN Brain Science Institute, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan. E-mail:


Mammals express two myo-inositol monophosphatase (IMPase) genes, IMPA1/Impa1 and IMPA2/Impa2. In this study, we compared the spatial expression patterns of the two IMPase gene transcripts and proteins in mouse tissues. Results indicated discrete expression of the two IMPase genes and their protein products in various organs, including the brain. In Caenorhabditis elegans, loss of the IMPase gene, ttx-7, disrupts cellular polarity in RIA neurons, eliciting abnormal thermotaxis behavior. We performed a rescue experiment in mutant nematodes using mammalian IMPases. Human IMPA2 rescued the abnormal behavioral phenotype in the ttx-7 mutants more efficiently than IMPA1. These results raise a question about the phylogenetic origin of IMPases and the biological roles of mammalian IMPase 2 in mammals. Impa2 knockout mice generated in our laboratory, exhibited neither behavioral abnormalities nor a significant reduction in myo-inositol content in the brain and other examined tissues. Given the ability of human IMPA2 to rescue the ttx-7 mutant, and its genetic association with multiple neuropsychiatric disorders, close scrutiny of IMPA2 function and the evolutionary origin of IMPase genes is warranted.

Abbreviations used

alkaline phosphatase


5-bromo-4-chloro-3-indolyl phosphate






enhanced green fluorescent protein


mammals express two myo-inositol monophosphatase


inositol trisphosphate




4-nitro blue tetrazolium


phosphatidylinositol 4,5-bisphosphate

Myo-inositol is an essential substrate for phosphatidylinositol synthesis in cells, as well as being an osmolite. In addition to direct uptake from extracellular spaces using the specific sodium/myo-inositol cotransporter (SMIT) and H+/myo-inositol cotransporter (HMIT) transporters, mammalian cells have two mechanisms by which they synthesize myo-inositol. The first is by recycling myo-inositol in a multistep dephosphorylation of inositol polyphosphate species, such as inositol trisphosphate (IP3), liberated from membrane-associated phosphatidylinositol 4,5-bisphosphate (PIP2). The second is through de novo synthesis from glucose phosphate, producing myo-inositol monophosphate. Importantly, the final step in both biochemical pathways, which is mediated by myo-inositol monophosphatase (IMPase: EC, is the dephosphorylation of myo-inositol monophosphate, producing phosphate-free myo-inositol. The mammalian IMPase enzymes, IMPase 1 and IMPase 2 (Yoshikawa et al. 1997), are encoded by the two separate genes, IMPA1 and IMPA2, respectively. While the primary structures of these two proteins are closely related to each other, we found that their 3D structure and enzymological features differed slightly (Arai et al. 2007; Ohnishi et al. 2007a; Fujita et al. 2011), suggesting a potential difference in biological roles between the two proteins.

It is thought that IMPase is one of the probable molecular targets for lithium salts (Berridge et al. 1989) used in the therapeutic treatment for psychiatric illnesses, particularly bipolar (manic-depressive) disorder. Furthermore, genetic variations in IMPA2, which encodes IMPase 2, have been implicated in multiple neuropsychiatric diseases, such as schizophrenia (Yoshikawa et al. 2001), bipolar disorder (Sjoholt et al. 2004; Ohnishi et al. 2007b), and febrile seizures (Nakayama et al. 2004). Lithium is known to inhibit IMPase activity both in vivo and in vitro, and its mood-stabilizing action is thought to be mediated through inositol depletion. These lines of evidence support crucial roles for IMPase enzymes in normal brain function as well as the pathogenesis of neuropsychiatric disorders.

In contrast to mammals, the nematode C. elegans has one IMPase gene, ttx-7, whose product shows a close structural homology to mammalian IMPase proteins. This set up makes C. elegans the simplest model for examining the biological consequences of a loss of IMPase in neurons. To date, a number of mutants with impaired associative learning-related behavior called thermotaxis, have been isolated in C. elegans (Mori et al. 2007). Of these, ttx-7 mutants exhibit mislocalization of post-/pre-synaptic proteins within neurites of the specific interneuron named RIA, resulting in a severely disrupted thermotaxis phenotype (Tanizawa et al. 2006). Recently, Kimata et al. (2012) revealed that mutations in egl-8 (encoding a phospholipase C β homolog) or unc-26 (encoding a homolog of synaptojanin 1 that dephosphorylates PIP2) suppress the ttx-7 deficiency-induced phenotype. Collectively, these data support the idea that myo-inositol monophosphatase directly regulates synaptic polarity by changing local concentrations of PIP2 on the cell surface of nematodes, as opposed to regulation through second messengers such as diacylglycerol (DG) and IP3. An attractive idea would be that the efficacy of lithium as a mood stabilizer is also mediated by modifying the cellular localization of synaptic proteins in the human brain.

In this study, we assessed the functional conservation of IMPase between vertebrates and invertebrates by examining whether human IMPases could rescue the neurological phenotype seen in nematode ttx-7 mutants (Tanizawa et al. 2006). We found that forced expression of human IMPase proteins rescued deficient neuronal polarity in nematodes. Interestingly and unexpectedly, the mutant phenotype was rescued more effectively by human IMPase 2 compared with human IMPase 1. We also created and analyzed Impa2 knockout (KO) mice to explore the biological role of IMPase 2 in vertebrates.

Materials and methods

Western blotting and in situ hybridization

Western blot analysis of IMPases in mouse tissue was performed following a standard procedure, using antibodies described elsewhere (Ohnishi et al. 2007a).

In situ hybridization using thin slices of the tissue was performed as described previously (Ohnishi et al. 2007a). In brief, digoxigenin (DIG)-labeled riboprobes for Impa1 and Impa2 were synthesized using a DIG RNA Labeling Kit (Roche Applied Sciences, Penzberg, Germany) and linearized pBluescript II KS+, containing either the mouse Impa1 or Impa2 open reading frame as a template (Ohnishi et al. 2007a). The probe was fragmented in an alkaline hydrolyzing solution (40 mM NaHCO3, 60 mM Na2CO3, pH 10.2) for 9.5 min at 60°C. After neutralization by adding 1/200x volume of acetic acid, the fragmented probe was precipitated by ethanol, and then checked by agarose gel electrophoresis.

Whole-mount in situ hybridization was performed as follows: embryos were quickly removed from pregnant females (C57BL/6N, Japan SLC, Hamamatsu, Japan) and treated in 6% H2O2/ethanol for 30 min, 75% ethanol for 5 min, and 50% ethanol/PBSTx (phosphate-buffered saline containing 0.1% Triton X-100) for 5 min, 25% ethanol/PBSTx for 5 min, PBSTx for 5 min, twice, and 10 μg/mL proteinase K/PBSTx for 15 min. After fixation in 0.2% glutaraldehyde/4% paraformaldehyde/PBSTx for 20 min and washing in PBSTx for 5 min twice, the embryos were treated in pre-hybridization buffer [50% formamide, 5x SSC, 5 mM EDTA, pH 8.0, 2% blocking reagent (Roche Applied Sciences), 0.1% Triton X-100, 0.1% CHAPS, 1 mg/mL yeast total RNA, 50 mg/mL heparin] at 62°C for 70 min. Hybridization was performed in a 400–800 ng/mL DIG-labeled probe solution overnight at 62°C. After extensive washings, embryos were treated with 1.5% blocking reagent in KTBTx [10 mM KCl, 150 mM NaCl, 500 mM Tris-HCl (pH 7.5), 0.1% Triton X-100] for 60 min, and then treated with alkaline phosphatase (AP) conjugated with anti-DIG antibody in blocking buffer (1 : 1000 dilution) for 3 h. Extensive washes in KTBTx and treatment with six washes of AP buffer [0.1 M Tris-HCl (pH 9.5), 50 mM MgCl2, 0.1 M NaCl, 0.1% Triton X-100, 5% polyvinyl alcohol] for 20 min were followed by color development in AP buffer containing 337.5 μg/mL 4-nitro blue tetrazolium (NBT) and 175 μg/mL 5-bromo-4-chloro-3-indolyl phosphate (BCIP). The reaction was stopped by extensive washing of the embryos in PBSTx, followed by treatment with 50% ethanol/PBTx for 5 min, absolute ethanol for 2 h, 50% ethanol/PBTx for 5 min, and PBTx for 5 min. To make the stained embryos transparent for better observation, they were soaked in 25% glycerol/PBSTx for 5 min followed by 50% glycerol/PBSTx for 5 min, and then stored in 50% glycerol until observation using a microscope (MZ16FA, Leica Microsystems, Wetzlar, Germany) equipped with a CCD camera (600CL, Pixera, Santa Clara, CA, USA).

Thermotaxis assay

Thermotaxis assays of Caenorhabditis elegans were performed according to the methods of Mori and Ohshima (1995) (Fig. 2c). In the rescue experiments, the open reading frames of human IMPA1 and IMPA2 were amplified by polymerase chain reaction (PCR) from SR-HA-IMPA1 and SR-HA-IMPA2 (Ohnishi et al. 2007a), respectively. Each fragment was inserted into an expression vector, containing an unc-14 pan-neuronal promoter (Ogura et al. 1997) and enhanced green fluorescent protein (EGFP) coding sequence to generate human IMPase::EGFP expression constructs, unc-14p::hIMPA1::EGFP (pTAN132) and unc-14p::hIMPA2::EGFP (pTAN133). These plasmids were injected into ttx-7 (nj50) mutant worms to generate extrachromosomal transgenic strains that were examined in behavioral rescue experiments. Detailed procedures for the rescue experiment are described elsewhere (Tanizawa et al. 2006). Expression of EGFP alone under the unc-14 promoter does not rescue the thermotaxis defect in ttx-7 mutants (data not shown).

Generation of Impa2 KO mice

All the procedures in this study were approved by the RIKEN Ethics Committee for Animal Study. KO mice generation (constructing targeting vector and obtaining targeted ES cells, chimeric mice, and F1 mice) was conducted by Ozgene Co. Ltd. (Bentley DC, Australia). The mice were generated on a C57BL/6 background. F1 mice were introduced to our institute (RIKEN Brain Science Institute) and crossed with CAG-FLPe Tg mice (Kanki et al. 2006),. which systemically express the FLPe recombinase. This enzyme promotes excision of the neo cassette, producing mice with a conditional KO allele, without a neo cassette. In this study, we did not use mice harboring a conditional allele. To obtain a systemic KO allele, the F1 mice with the neo cassette were crossed with CAG-Cre-Tg mice on a C57BL/6 background, resulting in mice heterozygous for the null allele. KO mice were maintained as heterozygotes by mating heterozygous males with inbred C57BL/6N females (Japan SLC, Hamamatsu, Japan). Resultant heterozygous females and males were intercrossed to produce wild-type mice, heterozygous, and homozygous (KO) mice for these experiments. Genotyping of mice was performed using multiplexed PCR, DNA samples prepared from tails, and primer mixtures of (KO_Fw: 5′-GACTGTGTTCTGGTTTCCTGTTTGGG-3′, KO_Rv: 5′-TGTT ATACTCCTGAACTTGTAGCAGC-3′, Impa2_Ex2 Fw: 5′-GTCAGAGTAAGGTCTGGTGGTTAAG-3′, and Impa2_Ex2Rv: 5′-GGA CCTGGGCTTGGACTTAACAACT-3′).

Tissue inositol levels

Tissue inositol content was determined according to an already described method (Berry et al. 2003), with slight modifications. Briefly, tissues were removed quickly from deeply anesthetized mice, frozen in liquid nitrogen, and stored at −80°C until use. On use, tissues were boiled for 10 min, in 500 μL of water containing 20 nmol [2d6] myo-inositol (CDN Isotopes, Pointe-Claire, Canada), to inactivate enzymes that could potentially hydrolyze inositol phosphate species. After repeated sonication with a microprobe (Misonix, Farmingdale, NY, USA), the extracts were further boiled for 10 min to complete the extraction of inositol. Extracts were centrifuged for 10 min and 200 μL of the supernatants were lyophilized, and then derivatized in 300 μL of a 1 : 1 mixture of pyridine and BSTFA/1% TMCS (Thermo Fisher Scientific, Waltham, MA, USA) for 12 h, at 65°C. We used a JEOL (Akishima, Japan) JMS-700V gas chromatograph mass spectrometer equipped with an electron ionization source in the high-resolution (= 5000) selected ion monitoring mode. The monitoring for myo-inositol and [2d6] myo-inositol were 612.4 m/z and 618.4 m/z, respectively. A standard curve for myo-inositol (Sigma-Aldrich, St. Louis, MO, USA) was generated for use in these experiments.

Behavioral analyses

In the hot-plate test, mice were placed individually on a plate (Model MK-350C, Muromachi-kikai, Tokyo, Japan) pre-heated to 52°C, and the latencies to licking of paws, flinching, and jumping were recorded manually. The cutoff time was set at 90 s. Other behavioral analyses were performed as described elsewhere (Watanabe et al. 2007; Ohnishi et al. 2010).

Statistical analysis

Data were analyzed by Student's t-test or Mann–Whitney U-test, using Prism software (GraphPad Software, La Jolla, CA, USA). < 0.05 was considered significant.


Expression patterns of Impa1 and Impa2 in various mouse tissues

The mammalian genome encodes two IMPase genes, Impa1/IMPA1 and Impa2/IMPA2. First, we compared the tissue distribution of these two gene products in mice to obtain insight into their functional differences and redundancies. Western blot analyses revealed a relatively broad expression pattern for Impa1 and varying expression levels for Impa2 in the tissues examined [Fig. 1a and (Ohnishi et al. 2007a)]. The Impa2 protein doublet (Ohnishi et al. 2007a) is enriched in salivary (submandibular) gland, tongue, thymus as well as kidney and digestive tracts [Fig. 1a and (Ohnishi et al. 2007a)]. Besides these tissues, other tissues including the brain show moderate expression of the Impa2 protein doublet. In contrast, Impa1 shows relatively broad expression in the examined tissues [Fig. 1a and (Ohnishi et al. 2007a)]. To obtain more detailed information on the spatial and developmental expression patterns of these two genes, we performed in situ hybridization on mouse kidney, tongue, and testis (Fig. 1b–d). The two genes appear to share a very similar mRNA expression pattern. Interestingly, we detected a very strong signal for Impa2 mRNA in both the epithelium and muscle layer of the tongue, while strong expression of Impa1 was restricted to the epithelium in the tongue. Next, we examined expression patterns of the two genes at embryonic time points. In E9.5 and E10.5 embryos, limb buds and the first brachial arch exhibited very high expression levels of both IMPase genes (Fig. 1e), possibly implicating IMPases in cell migration. No obvious differences were found in the spatial expression patterns between the two IMPase genes in the examined embryonic stages. We observed clear expression of mRNA from both IMPase genes in the forebrain (Fig. 1e), as well as other parts of the body, using whole-mount in situ hybridization.

Figure 1.

Distribution of the two mammals express two myo-inositol monophosphatase (IMPase) proteins in mouse tissue. (a) Tissue distribution of the two IMPase proteins. Lysates from various tissues were analyzed by western blotting with anti-Impa1 (top) and anti-Impa2 (bottom) antibodies. Arrowheads indicate the positions of specific signals. The asterisk shows non-specific anti Impa2 signal. (b–d) In situ hybridization. b; kidney, c; tongue, and d; testis. (e) Whole-mount in situ hybridization of E9.5 and E10.5 embryos. Results with antisense (left and right panels) and sense (middle panels) probes are shown.

Human IMPA2 rescues thermotaxis defects in the nematode ttx-7 mutant

C. elegans display an associative learning ability called thermotaxis, where animals cultivated with food at a certain temperature migrate back to this temperature when placed on a temperature gradient plate. The small set of characteristic neurons involved in executing this behavior has been identified as AFD, AIY, AIZ, RIA, RIB, and RIM (Mori et al. 2007). To dissect the genetic architecture of thermotaxis, a range of mutants deficient in the thermotaxis trait have been isolated. The ttx-7 mutant exhibits very specific and severe impairment because of mislocalization of post-/pre-synaptic proteins in RIA neurons (Tanizawa et al. 2006). These interneurons play a pivotal role in the integrity of the neuronal circuit for thermotaxis regulation. It is thought that ttx-7 is the only gene responsible for encoding IMPase in C. elegans (Fig. 2a and b). Next, we examined whether mammalian IMPases perform the same role in C. elegans, by assessing whether they could functionally rescue the C. elegans mutant (Fig. 2c). As shown in Fig. 2d, both human IMPases provided partial but significant rescue to the abnormal thermotaxis phenotype of ttx-7 (nj50 null allele) mutants. Interestingly, expressed human IMPase 2 rescued the phenotype to a greater extent than human IMPase 1 (Fig. 2d). This was unexpected, as a previous in vitro study had indicated that recombinant human IMPase 2 protein exhibits lower IMPase activity compared with human IMPase 1 (Ohnishi et al. 2007a). However, IMPA 2 shows genetic association with multiple neuropsychiatric diseases, including schizophrenia (Yoshikawa et al. 2001), bipolar disorder (Sjoholt et al. 2004; Ohnishi et al. 2007b), and febrile seizures (Nakayama et al. 2004). As mentioned before, lithium salts have long been the first-line drug for stabilizing moods in affected patients. Drawing together these lines of evidence, we speculated that mammalian IMPA2/Impa2 may have a specific and evolutionarily conserved function in the neuronal system.

Figure 2.

Human IMPA2 effectively rescues the nematode ttx-7 mutant. (a) Alignment of nematode, human, and mouse IMPase proteins (b) Phylogenetic tree, using the UPGMA method (, of nematode, human, and mouse IMPase proteins. (c) Schematic representation of the thermotaxis assay system. After exposure to food at 20°C, the animals were placed on a plate with a circular temperature gradient, and allowed to freely move for 1 h. (d) Rescue of ttx-7 mutants by human mammals express two myo-inositol monophosphatases (IMPases). Ttx-7 animals were injected with empty vector (promoter only), RIAp::hIMPA1::EGFP (human IMPA1), or RIAp:hIMPA2::EGFP (human IMPA2). Human IMPA2 rescued the ttx-7 phenotype more efficiently than human IMPA1 (< 0.01). *< 0.01 by Tukey's multiple comparison.

Impa2 KO mice grow normally without apparent phenotype

To test this premise in vertebrates, we generated and examined Impa2 KO mice. Targeting strategies for generating Impa2 knockout mice are illustrated in Fig. 3a. We analyzed mice with systemic KO alleles, although we also created mice with conditional KO alleles. In this case, exon 2 of the Impa2 gene is flanked by two loxP sites to generate the KO mice. Heterozygous KO females and males were intercrossed to produce wild-type (WT), heterozygote, and KO animals. Genotypes were determined by genomic PCR (Fig. 3b). Total absence of Impa2 mRNA in KO mice was confirmed by real-time RT-PCR (data not shown). Supporting this, no IMPase 2 protein (Ohnishi et al. 2007a) was detected in the tissues of KO mice (Fig. 3c, top panel). These data demonstrate that the KO mice had indeed lost their functional Impa2 gene, and that the observed protein doublet was a genuine product of the Impa2 gene. No compensatory up-regulation of Impa1 protein was found in various tissues from the KO mice (Fig. 3c, top panel).

Figure 3.

Impa2 knockout (KO) mice grow normally. (a) Targeting strategy for the generation of Impa2 KO mouse. Exon 2 (134 bp) was deleted in the KO alleles. Mice with conditional alleles were not analyzed in this study. (b) Genotypes of the mice (WT, HET, or KO) can be clearly determined by genomic PCR. A typical gel image after electrophoresis and ethidium bromide staining is shown. (c) Total loss of the Impa2 doublet in KO mice. Western blot analysis of various tissues from Impa2 KO mice was performed. Lysates of cerebral cortex, cerebellum, salivary gland (glandula submandibularis), kidney, thymus, and small intestine from WT and KO mice were examined with anti-Impa1 (upper panel) or anti-Impa2 (lower panel). Positions of relevant bands are indicated by arrowheads. (d) Comparison of face and body appearances of WT and KO mice. A typical photograph of the mice at 6 months of age is shown. There were no gross abnormalities in KO mouse. (e) Hydronephrosis-like abnormality in the kidneys of KO mice. Unilateral hydronephrosis (arrowhead) was observed in a small number of Impa2 KO mice. Typical photograph of right and left kidneys from WT and KO mice. Hydronephrosis was seen in about 4% of the KO mice, but did not meet statistical significance (by Fisher's exact test).

Homozygous KO mice were born within the Mendelian ratio, and grew normally (Fig. 3d) without any visible, abnormal phenotype. These results show that Impa2 is not necessary for normal growth in mice, while the Impa1 KO is lethal in the embryonic stage because of a relative decrease in tissue inositol content (Cryns et al. 2008). We detected no difference in the lifespan between Impa2 WT and KO mice (data not shown). A small proportion of the Impa2 KO mice showed kidney disease such as hydronephrosis (Fig. 3e). However, as the penetrance was very low (three affected kidneys from 74 KO mice), this finding was not statistically significant. The large proportion of Impa2 KO mice showed no functional or histological deficits in their kidneys or other IMPase 2-enriched tissues (Figure S1 and data not shown).

Next, we examined whether tissue inositol content altered between KO and wild-type mice. Results showed no changes in the brain regions, frontal cortex and cerebellum, or in the salivary gland, and thymus (Table 1). These findings contrast with those of Cryns et al. (2007) who reported reduced myo-inositol in the kidneys of their Impa2 KO mice. Collectively, we conclude that cellular myo-inositol levels, even in IMPase 2-enriched tissues, are maintained primarily by direct uptake, using the specific transporters, HMIT and SMIT, and/or, by dephosphorylation of myo-inositol monophosphate by IMPase 1.

Table 1. Inositol content of various tissues from Impa2 KO mice
Tissue Impa2 Student's t-test
WT (nmol/mg) (= 9)KO (nmol/mg) (= 8)p-value
  1. Data are shown as means ± SE.

Frontal cortex3.26 ± 0.0953.46 ± 0.0840.14
Cerebellum4.05 ± 0.154.11 ± 0.360.87
Salivary gland1.37 ± 0.111.31 ± 0.150.75
Thymus0.39 ± 0.0310.41 ± 0.0400.72
Kidney6.06 ± 0.335.59 ± 0.270.29

No behavioral phenotypes are evident in Impa2 KO mice

To explore potential abnormalities in the nervous system of Impa2 KO mice, we next conducted behavioral analyses using a test battery consisting of the open field, light–dark transition, pre-pulse inhibition (PPI), tail suspension, forced swim, home cage activity, passive avoidance, and hot plate tests (Tables 2 and 3, for male and female mice, respectively). As the human IMPA2 gene is associated with multiple neuropsychiatric diseases, including schizophrenia and bipolar disorder, we included a test paradigm for examining the emotional and cognitive domains. Although the mechanism is unknown, male KO mice exhibited significantly increased PPI (Table 2), an indicator of sensorimotor gating function, which is reduced in multiple neuropsychiatric diseases, including schizophrenia. No abnormality of PPI was observed in female mice (Table 3). Cryns et al. recently reported that their Impa2 KO mice showed moderately increased exploration behavior in the open field test (Cryns et al. 2007). However, we were not able to reproduce this finding (Tables 2 and 3). Results from the hot plate and passive avoidance tests, in which mice are subjected to electro-foot shocks, suggest that sensory neurotransmission from peripheral parts of the body to the brain, for example nociception, is also normal in KO mice. As explained earlier, IMPA2 is associated with febrile seizures (Nakayama et al. 2004). However, we found no evidence of seizures in KO mice under standard experimental conditions.

Table 2. Summary of behavioral tests on male Impa2 KO mice
TestIndexWT n KO n p (Student's t-test)
  1. All data are shown as mean ± SE.

  2. Bold letters indicate values that fulfilled statistical significance.

  3. a

    p by Two-way (genotype x pre-pulse level) repeated anova (genotype effect).

  4. b

    p by Mann–Whitney's U-test.

Open fieldTotal distance (cm)5917.5 ± 231.8225562.4 ± 288.3290.3647
Center time %24.2 ± 2.222.6 ± 1.60.5432
Elavated plus mazeOpen-arm stay %13.8 ± 4.92220.7 ± 4.3290.2943
Close-arm stay %63.3 ± 4.958.1 ± 4.50.4416
Center %22.9 ± 2.621.2 ± 2.00.5922
Light–dark transitionStay in light box (sec)109.5 ± 12.422136.6 ± 11.4290.1159
Stay in dark box (sec)505.9 ± 12.6476.9 ± 12.20.1100
Entry numbers16.3 ± 2.022.3 ± 1.9 0.0318
Latency (sec)179.8 ± 73.8104.3 ± 22.80.2809
Acoustic startle response (pre-pulse dB/pulse dB)PPI % (74/120) 14.0 ± 3.8 22 25.0 ± 2.6 29 0.0002 a
PPI % (80/120) 32.5 ± 4.7 49.8 ± 3.4
PPI % (86/120) 39.0 ± 5.5 59.4 ± 2.8
Pulse alone (arbitrary unit)1013.9 ± 79.3887.2 ± 58.50.1943
Tail suspensionImmobility %17.3 ± 2.42016.2 ± 2.1240.7182
Forced swimTotal immobility time (/5 min)152.7 ± 6.422152.3 ± 8.1290.9721
Home cage activity Total activity (arbitrary unit, x103/24 h)94.1 ± 8.62186.4 ± 7.0290.4844
Passive avoidanceLatency (sec, conditioning phase)48.2 ± 10.72271.2 ± 14.7290.4581b
Latency (sec, test phase)288.4 ± 10.1261.6 ± 16.70.2353b
Hot plateLicking (Latency, sec)27.9 ± 1.32127.5 ± 1.7270.873
Flinching (Latency, sec)21.9 ± 0.82223.0 ± 0.7290.9271
Jumping (Latency, sec)77.5 ± 6.51674.3 ± 5.0160.6392
Table 3. Summary of behavioral tests of female Impa2 KO mice
TestIndexWT n KO n p (Student's t-test)
  1. All data are shown as mean ± SE.

  2. a

    p by two-way (genotype × pre-pulse level) repeated anova (genotype effect).

  3. b

    p by Mann–Whitney's U-test.

Open fieldTotal distance (cm)2.905.6 ± 211.2222.532.1 ± 188.3240.1923
Center time %24.9 ± 4.828.0 ± 4.20.6261
Elavated plus mazeOpen-arm stay %21.9 ± 4.22228.2 ± 5.0240.3402
Closed-arm stay %55.7 ± 4.954.1 ± 5.00.8234
Center %22.4 ± 3.917.7 ± 1.70.2545
Light–dark transitionStay in light box (sec)105.5 ± 10.122122.1 ± 13.6240.3398
Stay in dark box (sec)495.7 ± 11.2483.7 ± 13.50.5031
Entry numbers20.8 ± 2.225.5 ± 2.90.2151
Latency (sec)86.3 ± 25.5112.1 ± 50.10.6568
Acoustic startle response (pre-pulse dB/pulse dB)PPI % (74/120)23.5 ± 5.12226.8 ± 3.4240.1131a
PPI % (80/120)34.2 ± 4.546.0 ± 4.6
PPI % (86/120)47.1 ± 3.855.0 ± 3.1
Pulse alone (arbitrary unit)677.0 ± 59.5649.5 ± 71.40.9512
Tail suspensionImmobility %22.3 ± 3.02020.9 ± 2.8230.7394
Forced swimTotal immobility time (/5 min)199.3 ± 7.022194.0 ± 6.9240.5929
Home cage activity Total activity (arbitrary unit, x103/24 h)150.8 ± 9.722155.5 ± 9.7240.7333
Passive avoidanceLatency (sec, conditioning phase)37.1 ± 7.42158.5 ± 12.6230.4447b
Latency (sec, test phase)264.7 ± 19.6282.2 ± 11.80.8132b
Hot plateLicking (Latency, sec)26.7 ± 1.22126.3 ± 1.4240.8107
Flinching (Latency, sec)23.7 ± 1.12122.8 ± 1.1230.5912
Jumping (Latency, sec)69.7 ± 5.81964.3 ± 3.8220.3018

As stated before, Impa2 is highly expressed in several tissues. But, the KO mice showed no morphological or histological abnormalities in these tissues (Figure S1). Taste sensing, which was analyzed by the two-bottle test (data not shown), appeared normal, despite high Impa2 expression in the tongue (Fig. 1a and c). In addition, KO mice showed no abnormalities in blood cells or biochemistry (data not shown). Collectively, our Impa2 KO mice exhibited no apparent phenotypes to suggest a biological function for the gene in mice.


In this study, we showed that forced expression of mammalian IMPase proteins in neurons of the C. elegans ttx-7 mutant, rescued their abnormal phenotype. The idea that expression of human IMPases reverted the ttx-7 phenotype by normalizing aberrant localization of the synaptic proteins in the mutant is strongly supported by following observations shown in our previous report (Tanizawa et al. 2006): (i) forced expression of functional ttx-7, the gene coding for a protein structurally homologous to mammalian IMPases, in RIA normalizes the thermotaxis defects and abnormal localization of the synaptic proteins in RIA from the ttx-7 mutant, (ii) the thermotaxis defects and abnormal localization of the synaptic proteins in RIA from the ttx-7 mutant can be alleviated by the supplementation of myo-inositol in the medium for the mutant, supporting that those phenotypes in the mutant are mediated by the depletion of cellular inositol, and (iii) an addition of lithium chloride, a potent inhibitor for IMPase, to the medium for the WT animals produces abnormalities, which are indistinguishable from those seen in the mutant, in thermotaxis and localization of the synaptic proteins. The ability of both human IMPase proteins to rescue the ttx-7 mutant phenotype supports that idea that both of the mammalian IMPases, IMPase 1 and IMPase 2, can function enzymatically as IMPases in vivo, even in C. elegans.

Interestingly, the potency to rescue the ttx-7 mutant phenotype was significantly higher for human IMPase 2 compared with IMPase 1. This was an unexpected observation as human IMPase 2 has much weaker IMPase activity than IMPase 1, at least in our in vitro assay conditions (Ohnishi et al. 2007a). We cannot rule out that this anomaly stems from the possibility that human IMPase 1 mRNA transcripts or protein levels may be less stable in nematodes than those of IMPase 2. This could be because of the existence of preferential binding partner(s) for human IMPase 2 in C. elegans or other unknown mechanisms. Both human IMPases were found to rescue the ttx-7 phenotype only partly, possibly because the native ttx-7 mRNA may retain cis element(s) to direct it to specialized sites in RIA as suggested in the mouse Impa1 transcript (Andreassi et al. 2010). These issues remain to be examined in the future study.

Based on these observations, it was attractive to postulate that mammalian IMPase 2 and invertebrate IMPase may have evolutionarily and functionally conserved role(s) in neurons. However, our extensive analyses failed to find any abnormal neurological or emotional phenotypes in Impa2 KO mice. In addition, no reduction of myo-inositol was observed in the brains of the KO mice. We have, however, identified an uncharacterized cell type that strongly expressed Impa2 in the midbrain to the hindbrain of mouse tissue (Ohnishi et al. 2007a). It would be of future interest to examine whether these cells show functional disturbance because of mislocalization of pre- and/or post-synaptic proteins in IMPA2 KO mice.

Human genetic association studies have implicated IMPA2 in the pathogenesis of multiple neuropsychiatric diseases. In particular, we have reported on a risk haplotype, which consists of multiple SNPs in the promoter region of IMPA2, leading to enhanced promoter activity of this gene, and additionally, this haplotype was overly expressed in post-mortem brains of patients suffering from bipolar disorder. These samples also showed a tendency for increased expression of IMPA2 mRNA compared with controls. (Ohnishi et al. 2007b). In this sense, it may be natural that our Impa2 KO mice, which lack the Impa2 mRNA and protein, showed no emotional disturbances.

It is likely that one of the biological differences between IMPase 1 and IMPase 2 resides in their binding partners. Berggard et al. reported that IMPase 1 binds to calbindin D28k, and that this modulates IMPase activity (Berggard et al. 2002). However, neither human IMPase 1 nor IMPase 2 associated with calbindin D28k in our coimmunoprecipitation assays (Arimoto et al., unpublished observation). Therefore, the interaction between IMPase and calbindin would be, if true, very weak or unstable in vivo. Chromatin immunoprecipitation assays followed by quantitative PCR (ChIP-qPCR) predict the existence of multiple transcription factor-binding sites in both IMPase genes. A putative ARNT (aryl hydrocarbon receptor nuclear translocator: a subunit of HIF-1, hypoxia-inducible factor 1)-binding site is located in the promoter region of Impa2/IMPA2, but not that of Impa1/IMPA1, potentially defining a functional specificity for each of the IMPase genes. However, IMPase 2 was not up-regulated under hypoxic conditions in HeLa cells (data not shown). Considering that the expression patterns of these two genes overlapped significantly in many tissues and embryonic time points, there seems to be functional redundancy between the two proteins. Thus, a double knockout of Impa1 and Impa2 in mice would be an important tool to examine why mammals have two IMPase genes in addition to specific transporters to supply cells with myo-inositol.

IMPA2 has been genetically implicated in the pathogenesis of bipolar disorder, schizophrenia, and febrile seizures as stated above. The nematode experimental system and our knockout mice could help to explore the biological function of the IMPA2 gene in the human brain.


We thank Drs. Itohara (RIKEN BSI) and Miyazaki (Osaka University) for kind donations of the CAG-FLPe mouse and CAG-Cre mouse, respectively. The authors are grateful to members of the Research Resources Center of RIKEN Brain Science Institute for technical support including, animal care, blood biochemistry, and blood cell counting, and Mr. Ishizuka (RIKEN BSI & BRC) for animal maintenance. This work was supported in part by Grant-in-Aids from MEXT of Japan (to T.O. and T.Y.), the RIKEN BSI Funds, and grants from the Mitsubishi Pharma Research Foundation (to T.Y.) and Takeda Science Foundation (to T.O.). In addition, a part of this study is the result of “Development of biomarker candidates for social behavior” carried out under the Strategic Research Program for Brain Sciences by the Ministry of Education, Culture, Sports, Science and Technology of Japan. The authors have no conflict of interest to declare.