New Spin on DISC1—Mouse Mutation Impairs Behavior
8 May 2007. In the May 3 Neuron, Steven Clapcote, David Porteous, and their colleagues report that simple missense mutations in the DISC1 gene lead to impaired behavioral function in mice. The results strengthen the case that variations in the gene play a role in complex neuropsychiatric diseases such as affective disorders and schizophrenia (see SRF live discussion).
DISC1, or disrupted in schizophrenia, was originally discovered by the Porteous group, at the University of Edinburgh, Scotland, when they found that a gross rearrangement of the gene associates with schizophrenia and affective disorders in a large Scottish family (see Millar et al., 2000). Since then, numerous studies have linked DISC1 polymorphisms with the disease in various populations (see SchizophreniaGene data); however, a robust cause-and-effect association has not emerged (see SchizophreniaGene meta-analysis). But this latest finding indicates that in mice, at least, DISC1 mutation is sufficient to cause some of the signs, or endophenotypes, commonly exhibited by people with the disease.
To find DISC1 mutations that might have a neurological impact, first author Clapcote, and colleagues in John Roder's group at Mount Sinai Hospital, Toronto, and elsewhere, screened progeny of mice that had been treated with the chemical mutagen N-ethyl-N-nitrosourea (ENU), which primarily causes simple DNA point mutations. The researchers identified the DISC1 mutations by sequencing exon 2 of the gene in the nearly 1,700 offspring. They focused on exon 2 because it is the largest DISC1 exon and it encodes the head domain of the protein, which interacts with phosphodiesterase 4B, a cyclic AMP degrading enzyme that the Porteous group has also linked to schizophrenia pathology (see SRF related news story). To ensure that other ENU-introduced mutations did not complicate the picture, the researchers back-crossed each mutant strain with the parental strain for five generations, which, according to the authors, should reduce the expected number of additional heterozygous mutations in each mutant line to 0.75.
To screen the mutant mice for neuropsychiatric changes, the authors tested them for prepulse and latent inhibition, two behaviors that are often compromised in patients with schizophrenia. In the former, startle responses elicited by a loud noise are suppressed by a prior stimulus that is insufficient to elicit the startle response. In latent inhibition, pre-exposure to a non-rewarding stimulus reduces subsequent interest in the stimulus even when it is later paired with a reward. Impairments in both are thought to be related to difficulties processing sensory information.
Clapcote and colleagues found that mice with either of two mutations—a leucine for a glutamine at position 31 (Q31L) and a proline for a leucine at position 100 (L100P)—were impaired in both prepulse (PPI) and latent inhibition (LI). The L100P mutation seemed to have a slightly stronger effect, with prepulse inhibiting the startle response in homozygous mutants only about 20 percent of the time, as opposed to 38 percent in Q31L homozygotes and about 60 percent in wild-type mice. Similarly, latent inhibition in L100P animals was about half that in Q31L mice, which in turn was only about one-third that in control mice.
In other behavioral tests, however, differences between the two mutants were not simply a matter of degree. The L100P animals appeared more anxious, moving around much more in an open field test, whereas the Q31L animals were not significantly different from wild-type. These animals also performed poorly in a T-maze test of working memory—both anxiety and impaired working memory are symptoms of schizophrenia. In contrast, the Q31L animals showed experimental behaviors believed to model severe depression. In a forced swim test they gave up more readily, floating for longer than controls and L100P animals, and they showed little interest in sucrose solution or strange mice introduced in a social interaction test.
The authors suggest that the Q31L mutation results in depressive-like behavior, while the L100P mutation exhibited schizophrenic-like behavior. In support of this dichotomy they found that while the antipsychotic drug clozapine restored PPI and LI somewhat in L100P animals, it had no effect on these parameters in Q31L mice. In the latter, PPI and LI impairments were relieved by the antidepressant bupropion, however.
How do these mutations cause such dramatic behavioral problems? The researchers found that both types of mice had significantly reduced brain volume—13 percent loss in the case of L100P and 6 percent for Q31L. These reductions were accompanied by contraction of the cortex, entorhinal cortex, thalamus, and cerebellum. These findings suggest that the mutation causes neurodevelopmental effects, an idea that has been previously postulated. But the researchers also found that levels of four major DISC1 isoforms in the brain appeared normal in the mutant animals, which suggests that the mutations do not affect gene expression but possibly alter activity of the protein.
Since Porteous's group previously showed that DISC1 binds with phosphodiesterase 4B (PDE4B), which may play a key role in regulating working memory (see SRF related news story), they chose to look at this interaction more closely. They found that both mutations reduced DISC1 binding to PDE4B when the proteins were expressed in cultured cells. However, they note that “the degree of binding was variable, particularly with the 31L mutation, suggesting that mutant DISC1 binding to PDE4B is influenced by the effects of unknown fluctuating cellular factors.” Nonetheless, the loss of PDE4B binding suggests that these mutations may impact cAMP signaling in addition to impairing neurodevelopment. “These mouse models support both the neurodevelopmental role for DISC1 and the proposed cAMP signaling role through modulation of PDE4 activity,” write the authors. Indeed, the Q31L mutants also had lower PDE4 activity and, unlike the L100P animals, the PDE4 rolipram had no effect on their PPI or LI.
When Is a Mouse a Model?
Indeed, this variability speaks to the very crux of the DISC1 matter. A clear picture of DISC1 biology and its influence on psychiatric illness has not emerged, possibly because DISC1 variations are responsible for many different phenotypes and are influenced by other genetic or environmental factors. “Our results in the mouse thus emphasize the importance of replicating, validating, and resolving inconsistencies within the current picture of various DISC1 alleles and haplotypes associating with distinct clinical phenotypes and, indeed, normal variation in cognitive function and neurodevelopment,” write the authors.
How valuable mouse models will be for complex human neuropsychiatric disorders remains to be seen. Mice, for example, do not have a prefrontal cortex, a region in the human brain that has been linked to psychotic behavior. And as Nancy Low, McGill University, Montreal, Quebec, and John Hardy, National Institute on Aging, Bethesda, Maryland, write in an accompanying Neuron preview, “Because psychiatric disorders are composed of lists of symptoms, many of which are nonspecific and overlap with those of other psychiatric disorders, a mouse model may be able to mimic some observable behaviors, but the attribution of a specific behavior to a human emotion or cognitive state cannot be made.”
Nonetheless, Low and Hardy suggest that, “These data are generally consistent with the view that genetic modification of the DISC1 locus leads to behavioral outcomes that may correspond to traits or abnormalities observed in psychiatric disorders.” However, they also urge caution in interpreting the data, given that the mutations “have complex biochemical, anatomical and behavioral effects.” They also conclude that “Mouse models, such as these, may be a way forward, but we must use them cautiously and be wary not to only study mice and lose sight of the human condition.”—Tom Fagan.
Clapcote SJ, Lipina TV, Millar JK, Mackie S, Christie S, Ogawa F, Lerch JP, Trimble K, Uchiyama M, Sakuraba Y, Kaneda H, Shiroishi T, Houslay MD, Henkelman RM, Sled JG, Gondo Y, Porteous DJ, Roder JC. Behavioral phenotypes of Disc1 missense mutations in mice. Neuron. 2007, May 3;54:387-402. Abstract
Low NC, Hardy J. What is a schizophrenic mouse? Neuron. 2007, May3;54:348-349. Abstract
Comments on News and Primary Papers
Comment by: Akira Sawa, SRF Advisor
Submitted 8 May 2007
Posted 8 May 2007
This is outstanding work reporting DISC1 genetically engineered mice.
Thus far, one type of DISC1 mutant mouse has been reported, by Gogos and colleagues
(Koike et al., 2006).
There are two remarkable points in this work. First, of most importance, John Roder and Steve Clapcote have been very successful in using mice with ENU-induced mutations for their questions. Due to the complexity of the DISC1 gene and isoforms, several groups, including ours, have tried but not succeeded in generating knockout mice. However, Roder and Clapcote found alternative mice that could be used in testing our main hypothesis. I believe that the majority of the success in this work is on this particular point. Indeed, to explore animal models for other susceptibility genes for major mental illnesses, this approach should be considered.
Second, it is very interesting that different mutations in the same gene display different types of phenotypes. I appreciate the excellence in the extensive behavioral assays in this work.
Although we need to wait for any molecular and mechanistic analyses of these mice in the future, this work provides us outstanding methodologies in studying major mental conditions. I anticipate that four to five papers will come out in this year that report various types of DISC1 genetically engineered mice. Neutral comparison of all the DISC1 mice from different groups will provide important insights for DISC1 and its role in major mental conditions.
View all comments by Akira SawaComment by: Christopher Ross
Submitted 8 May 2007
Posted 8 May 2007
This paper demonstrates that mutations in DISC1 can alter mouse behavior, brain structure, and biochemistry, consistent with the idea that DISC1 is related to major psychiatric disorders. This is already an important result. But more strikingly, the authors’ interpretation is that one mutation (L100P) causes a phenotype similar to schizophrenia, while the other mutation (Q31L) results in a phenotype similar to affective disorder.
There are a number of caveats that need to be considered. No patients with equivalent mutations have been identified. The behavioral tests have only a hypothesized or empiric relevance to behavior in the human illnesses. DISC1 itself, while a very strong candidate gene, is still not fully validated, and the best evidence for its role in schizophrenia still arises from the single large pedigree in Scotland.
Despite these caveats, I believe this paper is potentially a major advance. The authors’ interpretations are provocative, and could have far-reaching implications for understanding of the biological bases of psychiatric diseases. The models provide strong support for further study of DISC1. DISC1 has numerous very interesting interacting proteins and thus may provide an entry into pathogenic pathways for psychiatric diseases. We have suggested that interactors at the centrosome, involved with neuronal development, may be especially relevant to schizophrenia, while interactors at the synapse, or related to signal transduction, may be especially relevant to affective disorder (Ross et al., 2006). The beginnings of an allelic series of DISC1 mutations will presage more detailed genotype-phenotype studies in a variety of mouse models, with potential relevance to both schizophrenia and affective disorder.
View all comments by Christopher RossComment by: Nick Brandon (Disclosure)
Submitted 8 May 2007
Posted 8 May 2007
Mutant Mice Bring Further Excitement to the DISC1-PDE4 Arena
DISC1 continues to ride a wave of optimism as we look for real breakthroughs in the molecular events underlying major psychiatric disorders including schizophrenia, bipolar, and depression. In 2005, its fortunes became entwined with those of the phosphodiesterase PDE4B as they were shown to functionally and physically interact (Millar et al., 2005). Evidence linking PDE4B to depression has been known for some time, but in the wake of the DISC1 finding, its link to schizophrenia has hardened (Siuciak et al., 2007; Menniti et al., 2006; Pickard et al., 2007).
The Roder and Porteous labs have come together to produce a fantastic paper describing two ENU mutant mice lines with specific mutations in the N-terminus of DISC1. Luck was on their side as the mutations seem to have a direct impact on the interaction with the PDE4B. Furthermore, the two lines look to have distinct phenotypes—one a little schizophrenic, the other depressive. It is known from the clinical and genetic data that DISC1 is associated with schizophrenia, bipolar, and MDD, so this mouse dichotomy is very intriguing.
The mutant line Q31L is claimed to have a “depressive-like” phenotype. This comes from behavioral experiments including a range of assays looking at depressive-like behaviors where this strain had severe deficits, treatable with the dual serotonin-noradrenaline reuptake inhibitor (SNRI) bupropion, commonly prescribed for depression. Together these findings could just as easily be linked to the negative symptoms of schizophrenia. Furthermore, Q31L also shows modest deficits in two sensory processing paradigms (latent inhibition and pre-pulse inhibition), for which antipsychotics had no impact, and a working memory deficit, so this strain has characteristics of all the three key domains of schizophrenia. The pharmacology gets more interesting when these animals are dosed with rolipram (PDE4 inhibitor, raises cAMP levels) and look to be resistant to its effects. At the protein level, while it effects no changes in absolute levels of DISC1 and PDE4B, it leads to a 50 percent reduction in PDE4 activity. This information connects together nicely with the rolipram resistance, and thus the authors suggest that elevated cAMP might explain the behaviors observed, but they unfortunately do not show any cAMP levels in these animals. The paper also reports a decreased binding of the mutant form of DISC1 with PDE4B in overexpressed systems; coupled with the decreased PDE activity, this is in slight contradiction to the original Millar paper (Millar et al., 2005), but as the authors explain, the complexity of the DISC1-PDE4 molecular partnership could easily explain this. From my perspective, the lack of data to date on DISC1-PDE4 brain complexes is a major weak point of this story—this needs to be addressed as we move forward. This will also allow us to understand better the role of different DISC1 isoforms.
L100P is the “schizophrenic” brother of Q31P and has severe deficits in two sensory processing paradigms (latent inhibition and pre-pulse inhibition) which is reversed by typical and atypical antipsychotic and rolipram. Rolipram is able to modulate the behavior as PDE4 activity levels are at a wild-type level. Again, it shows decreased levels of DISC1-PDE4 binding.
Together, these two lines, along with the Gogos mice and a further bank of DISC1 mice which we should expect to see in the next year, puts the field in a position where we are now able to start to dissect out the clearly complex biological functions of DISC1. But as I indicated earlier, we need more information on relevant DISC1 isoforms. We know from the DISC1 interactome that there are many exciting partnerships to develop, but we may not have the fortune of an ENU screen to pull out mice with specific effects on an interaction. The differences in the behavior and pharmacology of these two strains is striking. In combination with the impact on PDE4-DISC1 binding and PDE4 activity, it highlights how much still needs to be understood for this interaction alone. More immediately, the mice show clearly that specific DISC1 mutations may give rise to specific clinical end-points and open up DISC1 pharmacogenomics as a real possibility.
View all comments by Nick Brandon
Comments on Related News
Related News: Working Memory—Adrenoreceptors and DISC1 in the Same cAMP?Comment by: Joseph Friedman
Submitted 11 May 2007
Posted 11 May 2007
Cognitive symptoms have emerged as an independent feature of schizophrenia that needs to be targeted for treatment independent of more well-known symptoms such as hallucinations and delusions. Indeed, the level of impairment in cognitive abilities is one of the strongest predictors of impaired adaptive life skills in patients with schizophrenia. The prefrontal cortex, critical for cognitive abilities such as working memory and executive functions, is well established to be dysfunctional in patients with schizophrenia. Although the significance of dopamine-related changes to the prefrontal cortex in schizophrenia has been extensively studied, noradrenergic changes are also important, but often overlooked. Moreover, second-generation antipsychotics, which partially address the reduced prefrontal dopamine activity in patients with schizophrenia, have only modest effects on the cognitive impairments associated with schizophrenia.
Alpha-2 noradrenergic agonists, such as the antihypertensive drug guanfacine, increase noradrenergic activity in the prefrontal cortex. Evidence demonstrating cognitive-enhancing effects of guanfacine on cognitive abilities related to the prefrontal cortex in both animals and healthy human subjects suggests a potential role for guanfacine in treating some of the cognitive impairments of schizophrenia. Although limited, there is some evidence in support of cognitive enhancing effects of guanfacine in patients with schizophrenia (Friedman et al., 2001). Our current clinical trial seeks to determine the reproducibility of these preliminary results and assess the potential effects of guanfacine on the adaptive life skills of patients with schizophrenia. This study is being conducted in the New York State Mental Health System, specifically at Pilgrim Psychiatric Center (Mount Sinai Hospital is the sponsor) and at the Bronx Veterans Administration Medical Center (the sponsor is the VISN3 MIRECC). We expect results by end of year 2008.
Should guanfacine be effective, my plan would be to try to obtain federal funding for a larger multi-site study of guanfacine in combination with either a social skills or cognitive skills rehabilitation program. However, even if proven effective, it is important to keep in mind that any potential guanfacine effects will be limited to cognitive abilities associated with the prefrontal cortex. As data from animal models and healthy human subjects indicate, guanfacine will most likely be ineffective in addressing important cognitive symptoms related to temporal lobe changes in schizophrenia.
View all comments by Joseph Friedman
Related News: Modeling Schizophrenia Phenotypes—DISC1 Transgenic Mouse Debuts
Comment by: David J. Porteous, SRF Advisor, Kirsty Millar
Submitted 2 August 2007
Posted 2 August 2007
Several genetic studies point to involvement of DISC1 in major psychiatric illness, including schizophrenia and bipolar disorder, but to date the only causal variant that has been definitively identified is the translocation between human chromosomes 1 and 11 that co-segregates with major mental illness in a large Scottish family and which directly disrupts the DISC1 gene (Millar at al., 2000). It has been speculated that a truncated form of DISC1 may be expressed from the translocated allele and, if so, that this could exert a dominant-negative effect, but there is no such evidence from studies of the translocation cases. Rather, the evidence from studies of lymphoblastoid cell lines carrying the translocation suggests that haploinsufficiency is the most likely disease mechanism in this family (Millar et al., 2005). The unresolvable caveat to this, of course, is that it has not been possible to determine whether this is true also for the brain. Moreover, it is far from certain that any productive product from the translocation chromosome would be a perfectly truncated protein encoded by all the remaining exons, as opposed to an exon-skip isoform, with or without a hybrid protein component borrowing sequence information from chromosome 11. What does seem likely from other human studies is that additional genetic mechanisms, including missense mutations, altered expression, and possibly also copy number variation, play a role in the generality of DISC1 as a risk factor.
The evidence in support of DISC1 as an important biological determinant across a spectrum of major mental illness has now received a further boost from the study in PNAS by Hikida et al. The Sawa lab describes a transgenic approach where a truncated human DISC1 protein is expressed from a CAMKII promoter. The truncation was designed to mimic the hypothetical truncation arising from the Scottish translocation. This forebrain-specific promoter confers preferential expression of the transgene at neonatal stages, as distinct from the expression of the endogenous protein, which is abundant from embryonic development into adulthood. This model therefore permits investigation of the effect of the truncated protein in the forebrain within a specific developmental window, against a background of endogenous mouse DISC1 expression. Since there is no evidence for production of a truncated protein from the translocated allele, the relevance of this model to psychiatric illness remains unclear. However, on the positive side and from a functional perspective, dominant-negative effects as a consequence of expressing the truncated protein have already been demonstrated in cultured cells, altering the subcellular distribution of DISC1 and interaction with DISC1 partner proteins. Moreover, expression of the truncated form of DISC1 mimics downregulation of DISC1 in vivo, inhibiting migration of neurons in the developing mouse cortex (Kamiya et al., 2005). Thus, this model has the genuine potential to enhance our understanding of the biology of DISC1.
This is, in fact, the third study describing mice expressing modified DISC1 alleles. In the first study, Gogos and colleagues (Kioke et al., 2006) studied the effects of a modified DISC1 allele carrying a spontaneous 25 bp deletion in exon 6 that is present in all 129 mouse strains (Koike et al., 2007; see SRF related news story). This allele additionally has an artificial stop codon in exon 8 and a downstream polyadenylation signal. After back-crossing this mutagenised version of the 129 allele onto a C57Bl6 background, they reported a deficit in an assay of working memory in both heterozygous and homozygous mutants, but other standard behavioral tests were unaltered or unreported, and there were no anatomical, electrophysiological, or pharmacological studies included. In the second study, one led by the Roder laboratory, Toronto, we described two mouse strains with missense mutations in exon 2, Q31L and L100P (Clapcote et al., 2007). Reductions in brain volume, deficits in a range of standard behavioral tests, and responses to pharmacological treatments were reported, which can be summarized as consistent with the 100P mutants displaying schizophrenia-like phenotypes and the 31L mutants, mood disorder-like phenotypes. There is a frustrating dearth of consistent biomarkers for schizophrenia, but one of the best replicated findings is by brain imaging of enlarged ventricles in schizophrenia (also, but less markedly, in bipolar disorder). Adding to the observations of Clapcote et al., arguably the most striking claim by Hikida et al. is for altered ventricular brain volume and reduced brain laterality following neonatal transgenic overexpression of truncated DISC1. Additionally, behavioral tests were reported that overlap in part with those reported earlier by Clapcote et al. That three studies all describe behavioral abnormalities consistent with modeling components of schizophrenia is reassuring. That there are clear differences, too, between the phenotypes should come as no surprise either, given the important differences in terms of genetic lesion and developmental expression. Other mouse models are in the pipeline and they, too, will be welcomed. Indeed, this is very much what is needed for the field to move forward. But we should do so with some caution, paying careful attention to the specific nature of the models, what they can and cannot tell us about DISC1 biology, and what they may or may not tell us about the human condition. Although none of these models relates directly to a known causal variant, it appears that the mouse models concur with the human genetic studies in suggesting that there are likely to be several routes by which DISC1 can perturb brain function leading to characteristics of human mental illness. What we need now is for the human genetic studies to catch up with the mouse so that defined pathognomic variants can be modeled.
View all comments by David J. Porteous
View all comments by Kirsty Millar
Related News: Modeling Schizophrenia Phenotypes—DISC1 Transgenic Mouse Debuts
Comment by: John Roder
Submitted 2 August 2007
Posted 2 August 2007
A new mouse model from the Sawa lab strengthens the evidence for the candidate gene DISC1 playing a role in psychosis and mood disorders. This important paper is the first to address one potential disease mechanism, that of a dominant-negative effect. Expression of the C-terminal deletion of human DISC1—which represented the original rearrangement found by the Porteous group in the Scottish families with schizophrenia and depression—in transgenic mice driven by the α CaMKII promoter, first described by Mark Mayford when a postdoctoral fellow in the Kandel lab, leads to mice showing behaviors consistent with schizophrenia and depression, with enlarged lateral ventricles. Since the Sawa group expressed the human C-terminal truncation in mouse with no change in mouse DISC1 levels, they feel this supports a dominant-negative mechanism. More direct experiments are required. For example, create a null mutant mouse for DISC1 and express the full-length and truncated human DISC1 under the influence of their own promoter in transgenic mice using human BACs. Full-length human DISC1 would, hopefully, rescue the null. If so, then a mixture of full-length and truncated DISC1 proteins could be tried. No rescue by the mixture of full-length and truncated proteins would suggest a dominant-negative mechanism.
The Porteous group has shown no detectable DISC1 protein in lymphoblasts from the patients, and put forward the following implicit model. The C-terminal truncation of DISC1 makes the protein unstable and sensitive to degradation, a plausible alternative notion. In my opinion both are likely in different schizophrenia patients with perturbations in DISC1, most of which are alterations other than the C-terminal truncation. Some have SNPs that lead to as yet uncharacterized disease. It has been shown by the Sawa lab that mice with a partial RNAi knockdown of DISC1 show perturbations in brain development, and if aged to 8-12 weeks these mice might have shown behavioral and neuropathological hallmarks of schizophrenia and depression. There is currently no null mutation in the mouse that would address this issue, although DISC1 is one of the genes being targeted in the NIH knockout mouse project. Fortunately, there are now several mouse models—the more the better. The Gogos lab has a 25bp deletion in exon 6 that removes some, but not all isoforms. The Roder lab used a reverse genetic screen of an ENU archive to generate two missense mutants in non-conserved amino acids. The phenotypes of all these lines are nicely summarized in the Sawa paper. This work represents a step forward in our understanding of the DISC1 gene.
View all comments by John Roder
Related News: The DISC1 Switch in Neurodevelopment
Comment by: Albert H. C. Wong
Submitted 13 May 2011
Posted 13 May 2011
This recent and important paper by Sawa's group adds another layer to the complex story of DISC1 function in neurodevelopment. Their findings clarify and integrate two streams of research implicating DISC1 in both neuron proliferation and migration. The identification of the S170 phosphorylation site also raises the exciting possibility that pharmacological strategies targeted at this phosphorylation-dependent switch might be useful in correcting or preventing mental illness-related problems with brain development. It would be interesting in this context to explore whether disease-associated DISC1 gene variants in humans affect DISC1 phosphorylation, and the subsequent balance between neuron proliferation and migration.
I agree with Atsushi Kamiya that further work is needed to understand which of the many effects of DISC1 perturbation are specific to human psychiatric disease phenotypes. Again, from a treatment perspective, it is vital to know which cellular abnormality underlies the most debilitating symptoms so that new treatments can be screened for effects on these specific abnormalities. Another recent paper from our group reinforces this point (Lee et al., 2011). We found that genetic inactivation of GSK3α restored dendritic spine deficits in DISC1 L100P mutant mice, in parallel with amelioration of behavioral abnormalities as previously reported (Lipina et al., 2011). However, other abnormalities in dendrite morphology caused by the DISC1 L100P mutation were not corrected by GSK3α inactivation.
Lee FH, Kaidanovich-Beilin O, Roder JC, Woodgett JR, Wong AH. Genetic inactivation of GSK3α rescues spine deficits in Disc1-L100P mutant mice, Schizophrenia Research. 2011;Apr 16. Abstract
Lipina TV, Kaidanovich-Beilin O, Patel S, Wang M, Clapcote SJ, Liu F, Woodgett JR, Roder JC. Genetic and pharmacological evidence for schizophrenia-related Disc1 interaction with GSK-3. Synapse. 2011;65:234-248. Abstract
View all comments by Albert H. C. Wong