Normalizing Endophenotypes of Schizophrenia: The Dip and Draw Hypothesis

Saab et al, Graphical Abstract

Schizophrenia is a multifactorial complex genetic disorder generally characterized by a copious and polarized array of features that cause an anomalous perception of reality and social dysfunction. Timely and accurate diagnosis of schizophrenia can be obscured due to comorbidity and treatment is often unsatisfactory. Two models, the dopamine hypothesis and glutamate hypothesis, attempt to explain the underlying mechanisms of the disease. Importantly, the hypotheses are not mutually exclusive and may work together in the manifestation of schizophrenia, each playing an independent role for a subset of symptoms. Finding causes of the disease has been extremely difficult, largely due to its phenomic complexity. As a consequence, psychiatrists have begun to document endophenotypes, quantifiable symptoms with a molecular basis, for the disease in attempts to deconstruct, simplify and focus schizophrenia research. Endophenotypes can be present in model organisms, allowing for elegant and controlled experimentation. A recently generated reversible animal model of aberrant dopaminergic activity provides support for a novel understanding of how the brain may respond to antipsychotics. A prediction herein named the “dip and draw hypothesis” is presented to explain discrepancies between the early and delayed-onset hypotheses of antipsychotic action. Although progress in schizophrenia research has been modest over the last century, the recent union of theory and technology may provide the potential for better treatment.

Schizophrenias: Splitting of the Mind
Schizophrenia is a multifactorial complex genetic disorder affecting close to 1% of the world’s population, or about fifty million people (1). It is generally characterized by a combination of several features including disorganized thinking, delusions, hallucinations, inappropriate emotional responses, sensorimotor defects, withdrawal and loss of motivation. These symptoms summate in an anomalous perception of reality and social dysfunction. As a result, unemployment and poverty are common among people with schizophrenia. Because of the prevalence of schizophrenia and its impact on afflicted individuals and their families, it has been under considerable investigation since the condition was first described by Emil Kraepelin in 1893. At this time, schizophrenia was entitled “dementia praecox“, meaning “premature dementia,” because it was detected in young adults in contrast to most forms of known dementia, which began later in life. However, when another Swiss psychiatrist, Eugen Bleuler, noticed the prognosis of his patients often improved, he concluded the disorder could not be accurately described as dementia. As an alternative, he proposed the term, “schizophrenias” (Greek for “split” or “divided brain”) which remains today in the singular form. Bleuler intentionally introduced the term in the pluralized form to emphasize the diverse combination of symptoms manifest in people with schizophrenia.

In addition to naming the disorder, Bleuler divided the multiple symptoms of schizophrenia into two categories: positive and negative. Positive symptoms are thoughts, perceptions and behaviours normally absent in non-affected people. They include hallucinations, delusions and thought disturbances. Negative symptoms, on the other hand, defined by the absence of thoughts, perceptions and behaviours normally present in non-affected people, include apathy, inattention, anhedonia, and diminished facial and tonal gestures such as scarce, ineffective speech. The copious and polarized array of schizophrenic symptoms also has a high degree of comorbidity with other disorders, like clinical depression and generalized anxiety. As a result, timely and accurate diagnosis is often obscured when only a fraction of these symptoms are expressed. Though the cause of schizophrenia is still poorly understood, there exists a clear consensus that the disease is multifactorial in origin; with contributions from environmental, genetic, and epigenetic elements.

The Cause and the Cure
Two major hypotheses aim to explain the molecular mechanisms underlying the symptoms of schizophrenia. The first centers on aberrant dopamine transmission, while the second centers on aberrant glutamate transmission. The dopamine hypothesis arose primarily from the observation that antipsychotics used to treat schizophrenia antagonize dopamine type-2 receptors (D2Rs). Strength for this hypothesis also comes from brain imaging data showing various abnormalities within the dopaminergic system of drug-naïve patients with schizophrenia. For example, D1Rs are overexpressed in the striatum (2), a region associated with the automatic processing of information (3), while D2Rs are hypofunctional in the right thalamus (4), a region responsible for relaying sensory and emotional information to the cortex (5). The glutamate hypothesis is based on the well understood role for glutamate in all types of cognition and on evidence that administration of blockers of glutamate transmission can cause human and experimental animals to adopt some of the negative symptoms of schizophrenia. Furthermore, clinical trials with glutamate receptor agonists targeted to the N-methyl-D-aspartate receptor have shown some promise. Hypofunction of glutamenergic activity is therefore a viable model for schizophrenia (6). Importantly, the hypotheses are not mutually exclusive; they could work together in the manifestation of schizophrenia, each playing an independent role for a subset of symptoms.

Endophenotypes: Strength by Division
Irrespective of the molecular mechanism, finding genetic, environmental and epigenetic causes of the disease has been extremely difficult, largely due to its phenomic complexity. As a consequence, psychiatrists have begun to document “endophenotypes” for the disease in attempts to deconstruct, simplify and focus schizophrenia research. Endophenotypes are quantifiable symptoms with a molecular basis that can be present in non-affected siblings and are unseen by the unaided eye (7). Originally, the term was invented to explain concepts in insect biology (8), but its application was soon expanded to include psychiatric disorders (9, 10). Disorganized thought and sensorimotor gating defects are each considered endophenotypes of schizophrenia.

The advantage of using endophenotypes to describe a disease instead of overt generalized symptoms lies in how endophenotypes can be studied. First, endophenotypes can be present in humans who are not affected by the core disease and can therefore be studied in healthy subjects. Studying healthy individuals reduces any effects of disease-treatment on the endophenotype (i.e. helps address the cause versus consequence dilemma). The display of endophenotypes in related family members is also important since it alters the outcome of linkage analysis. For example, because siblings of people with schizophrenia also have sensorimotor defects (7, 11, 12), the pool of genetic material associated with this endophenotype is distinct from the pool of genetic material associated with the disease. Some investigators have found the use of endophenotypes modestly improves the power of schizophrenia linkage analysis (7, 13). (Of course, this does not mean endophenotypes of diseases necessarily have a simpler genetic architecture (14).) Second, because endophenotypes are not necessarily associated with the rout disease, they can be induced and measured in model organisms that otherwise have no real potential for recapitulating the pathology in an experimental setting. Some endophenotypes of schizophrenia, such as sensorimotor defects and disorganized thought are reproducible in animals and are therefore amenable to controlled and elegant investigation not possible in humans.

Animal Models of Schizophrenic Endophenotypes
Early animal models of schizophrenia were generated by injecting rats with amphetamine or phenylethylamine, both stimulators of dopaminergic activity (15, 16). More recently, administration of glutamate receptor blockers, such as MK-801 or phenylcycladine (PCP), has also been found to generate endophenotypes for schizophrenia in animals (17, 18). Animal models have also been generated via lesions to brain areas thought to be important for the endophenotypes of the disease (19). For example, early-life lesion to the hippocampus, a brain area important for memory formation and spatial awareness, produces sensorimotor deficits in adult mice and rats (20). Yet, because drastic measures are required to induce these endophenotypes (particular lesions), it is hard to imagine how relevant they are for understanding and developing therapeutic approaches to schizophrenia. Modest genetic alteration or environmental insult, on the other hand would recapitulate the manner of schizophrenia onset much more closely.

While animal models of schizophrenia created solely by environmental factors have yet to emerge, a handful of recently generated mutant mice express endophenotypes of schizophrenia. Clapcote and Lipina et al. (21) discovered two lines of mice each with a point mutation in the gene disrupted in schizophrenia 1 (disc 1). The mice display several endophenotypes of schizophrenia including sensorimotor defects and impaired working memory. Others have generated knock-out mice that display sensorimotor gating abnormalities, such as for zic 1 and zic 2 genes (22, reviewed in 23), or other endophenotypes of schizophrenia (24-29). Similarly, Kellendonk et al. (30) recently investigated the relationship between dopamine type-2 receptors and schizophrenia endophenotypes. The model created by Kellendonk et al. brings to light intriguing features of the endophenotypes that can only be visualized through the elegant reversibility of their system. It is worth considering their experiments in more detail.

Reversing Endophenotypes of Schizophrenia
Kellendonk et al. used a genetic approach to model the altered dopaminergic activity found in the striatum of untreated patients with schizophrenia (2, reviewed in 31). Using the tetracycline system, the authors overexpressed human D2Rs selectively in the striatum of double transgenic mice. D2R overexpressing double transgenic animals (referred to as D2R mice or D2R animals hereafter) demonstrated difficulties in organized thinking when compared to their littermates. Interestingly, defects in sensorimotor gating were not found, suggesting altered dopaminergic signaling in the striatum may not underlie this endophenotype in mouse.

D2R mice made more errors and took longer to reach criterion in a radial arm maze, a test specifically designed to measure working memory, than did their single transgenic or wild-type litter mates. D2R mice were also impaired in the T-maze, a separate measure of working memory in rodents. Further to working memory, D2R mice demonstrated impaired mental flexibility as measured by a reversal-learning odor-discrimination task. Together, these results demonstrate dopaminergic signaling in the striatum is important for organized thinking and may underscore some pathological aspects of schizophrenia in human.

To discern if the observed schizophrenic endophenotype was due to concurrent overexpression of the D2 receptor and not chronic or developmental effects, the authors switched off the transgene during adulthood (at ~3 months of age). Remarkably, the deficits in working memory, as assessed by the T-maze, persisted even after the level of D2Rs in the striatum normalized. Thus, it is likely the defect was rendered via a developmental or chronic mechanism. From this finding, the authors speculate reducing D2R levels in the striatum by the use of antipsychotics (which all antagonize D2Rs) may not be an effective form of therapy in human. However, the D2R animals were only relieved of striatal overexpression for a relatively short period (2 weeks). For reasons discussed below, it would be interesting to re-examine the endophenotype after longer periods of relief.

The irreversibility of the endophenotype confronted Kellendonk et al. with an interesting question: How could overexpression of a receptor in one brain region lead to behavioural deficits normally attributed to dysfunction in a distinct and distant brain region? Typically, working memory in mammals, including human and rodent, is attributed to prefrontal cortex (PFC) function (32). Therefore, one should expect the D2R mice to have defects in PFC function, in addition to defects in striatal function.

To tackle this hypothesis, Kellendonk et al. measured global levels of dopamine and dopamine-type 1 receptor activity as markers for dopaminergic function within the PFC. They saw a two-fold elevation in dopamine and enhanced D1R activity. Whole cell D1R expression levels were unaffected. The authors propose the effects on dopaminergic PFC function are mediated through a loop of brain circuitry known as the cortico-striatal-pallido-thalamo-cortical associative loop that connects the striatum to the cortex (Figure 1). More specifically, they propose either PFC D1R surface expression or receptor sensitivity is increased during development as a consequence of D2R overexpression in the striatum.

Figure 1: The cortico-striatal-pallido-thalamo-cortical loop. In this cartoon of a bisected human brainhemisphere, the connections between striatum,pallidum, thalamus and cortex are shownschematically. Through this loop,compensation for alterations in one region canbe induced in distant structures. Th =Thalamus, Pal = Pallidum, Str = Striatum, D =dorsal, V = ventral, M = medial, L = lateral, A =anterior, P = posterior.

If PFC D1R hyperfunction is responsible for the endophenotype in the D2R animals and turning off D2R overexpression does not rescue the endophenotype, then turning off D2R overexpression should also not rescue PFC hyperfunction. Yet, when Kellendonk et al. measured PFC D1R activity in the D2R mice after 2 weeks of endogenous D2R expression, they discovered that the hyperfunction does not persist. In fact, PFC D1R activity was below normal, demonstrating that hypofunction as well as hyperfunction of D1Rs in the prefrontal cortex can lead to impairment in working memory and mental flexibility. This observed U-shaped relationship between D1R activation and working memory has been documented before (33). The discovery also suggests that changes in D2R levels in the striatum can induce reversible adjustments in D1R activation in the prefrontal cortex. To explain this reversibility, a simple homeostatic hypothesis is suggested whereby striatal dopaminergic activity regulates prefrontal cortical dopaminergic activity (Figure 2). Following normalization of D2Rs, activity in the PFC over-compensates and dips below ‘normal’. Thus, the endophenotype persists even after D2R expression in the striatum returns to endogenous control.

Figure 2: Hyperactivity of D1R in the Prefrontal Cortex. Overexpression of D2R in the striatum results in hyperactivity of D1R in the prefrontal cortex (PFC).When the expression of D2R is normalized, D1R activity in the PFC overcompensates to a level belownormal. The shaded region represents pathological levels of PFC D1R activity. Closed circles = PFCD1R, open circles = striatal D2R.

Antipsychotics: The Dip and Draw Hypothesis
Given the striking ability for the cortico-striatal-pallido-thalamo-cortical loop to undergo plasticity, it is attractive to hypothesize that normalizing dopaminergic activity in one area of the loop might eventually lead to normalization of dopaminergic activity in another area of the loop. In the experiments conducted by Kellendonk et al., D2R animals were only relieved of striatal overexpression for a relatively short period. It would be interesting to re-examine both the endophenotype and PFC D1R activity after long periods of relief to see if they continue to persist. Potentially, prefrontal cortical dopaminergic activity would normalize and the endophenotype would be reversed (Figure 3).

Figure 3: The dip and draw hypothesis of antipsychotic action. D1R activity in the PFC re-normalizes to 100% after a prolonged period of time. Closed circles = PFC D1R, open circles = striatal D2R, closed circles with white border = PFC D1R predicted by the dip and draw hypothesis.

This type of “dip and draw” phenomenon where normalization is preceded by overcompensation is widespread in nature [ex. see ref. 34-36 and reviewed in 37]. The concept can be modeled by visualizing two buckets of water (a and b) connected through a pipe (p) at their bases, with bucket a representing the striatum, bucket b, the prefrontal cortex, and pipe p, the cortico-striatal-pallido-thalamo-cortical loop. When the level of water in bucket a is slowly increased, the level of water in bucket b will also slowly increase as the water is drawn between them through the pipe. Similarly, if water is slowly removed from bucket a, the same change will occur in b. However, if water is abruptly removed from one bucket, the level in the other will at first drop below the adaptive level due to the momentum of the water through the pipe. Only after the momentum is dissipated will each bucket normalize at the proper volume. Thus, relieving striatal overexpression in the D2R mice is, by analogy, an abrupt volume change in bucket a (the striatum) causing a transient “dip” in dopaminergic activity in bucket b (the prefrontal cortex). Eventually, however, the dip in dopaminergic activity will “draw” to a normalized level.

An Explanation to Consolidate the Early and Delayed-Onset Hypotheses
The classical understanding of antipsychotics pharmacology includes the belief that they have a delayed-onset of action; that is, they only start working several weeks following initial administration (38). While this view correctly describes overt expression of psychosis, the related assumption, that no change in disease outcome is seen immediately due to some unknown slow-acting adaptation mechanism of the brain, has recently been challenged. Analyses that pool results from multiple clinical trials demonstrate that an early-onset hypothesis is equally, if not more, valid (39-41).

One of the strengths for the dip and draw hypothesis is that it can reconcile the discrepancy between the delayed and early hypotheses. If the administration of antipsychotics is viewed as an abrupt change in striatal D2R levels (as in the D2R mice), then prefrontal cortical D1R activity might undergo adaptive regulation via the cortico-striatal-pallido-thalamo-cortical loop as proposed for the D2R animals. Thus, an effect of the antipsychotics could be seen at two time points: First when D1R activity approaches but passes the proper level and second when D1R activity stabilizes correctly (Figure 3). Unless these processes occur with exactly parallel time courses in every treated individual, the response to any antipsychotic will appear extremely variable initially and only later stabilize when the drug begins “working” after a delayed period. Bearing in mind the data used to generate the early-onset hypothesis are averaged from hundreds of individual reports, it is entirely conceivable that the treatment-effect slope observed is actually the result of benefits and over-compensation occurring at different times in different patients (Figure 4).

Figure 4: The Early-Onset Hypothesis. The early-onset hypothesis for antipsychotics assumes pharmacological agents as treatments forschizophrenia have immediate and graded effects. This observation could be explained by averagingthe response profile of multiple patients, each bi-directionally affected by the antipsychotics. The dipand draw hypothesis predicts a region of high variability immediately following antipsychoticadministration (treatment). Thin lines = individual patients, grey circles = observed slope for all patients.

While schizophrenia has been difficult to understand and treat due to its multifactorial nature, novel theoretical approaches such as the endophenotype dissection and emerging technologies such as tissue restrictive and inducible transgenics are coming ever closer to finding answers.

Glossary of Key Terms:
Anhedonia: Inability to feel pleasure
Endophenotype: Quantifiable symptoms of a disease with a molecular basis.
Schizophrenia: A mental disordercharacterized by some, but not necessarily all, of the following features: emotional blunting, intellectual deterioration, social isolation, disorganized speech and behaviour, delusions, and hallucinations.
U-Shaped Relationships: Refers to situations in which two extremes result in similar outcomes, and these outcomes are distinct from the mean of these extremes.


1. C. A. Ross, R. L. Margolis, S. A. Reading, M. Pletnikov, J. T. Coyle.T Neuron 52, 139-53 (2006).

2. D. F. Wong, H. N. Jr. Wagner, L. E. Tune, R. F. Dannals, G. D. Pearlson et al. Science 234, 1558-63 (1986).

3. B. J. Knowlton, J. A. Mangels, L. R. Squire. A neostriatal habit learning system in humans. Science 273, 1399-402 (1996).

4. M. Talvik, A. L. Nordstrom, Y. Okubo, H. Olsson, J. Borg et al. Psychiatry Res. 148, 165-73 (2006).

5. S. M. Sherman, R. W. Guillery. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 357, 1695-708 (2002).

6. J. T. Coyle. Cell. Mol. Neurobiol. 26, 365-84 (2006).

7. I. I. Gottesman, T. D. Gould. Am J Psychiatry 160, 636-45 (2003).

8. B. John, K. R. Lewis. Science 152, 711-721 (1966).

9. I. I. Gottesman, J. Shields. Schizophrenia and Genetics: A Twin Study Vantage Point. (Avademic Press), New York, (1972).

10. I. I. Gottesman, J. Shields. Br. J. Psychiatry 122, 15-30 (1973).

11. D. Wong, M. Maybery, D. V. Bishop, A. Maley, J. Hallmayer. Genes Brain. Behav. 5, 561-76 (2006).

12. S. Bolte, F. Poustka. Psychol. Med. 33, 907-15 (2003).

13. Y. L. Liu, C. S. Fann, C. M. Liu, C. C. Chang, W. C. Yang et al. Mol. Psychiatry (2007); published online Mar 6.

14. J. Flint, M. R. Munafo. Psychol. Med. 37, 163-80 (2007).

15. J. D. Gambill, C. Kornetsky. Psychopharmacology 50, 215-23 (1976).

16. R. L. Borison, H. S. Havdala, B. I. Diamond. Life Sci. 21, 117-22 (1977).

17. P. I. Tiedtke, C. Bischoff, W. J. Schmidt. J Neural Transm. Gen. Sect. 81, 173-82 (1990).

18. M. F. Piercey, C. A. Ray. Life Sci. 43, 379-85 (1988).

19. B. K. Lipska, G. E. Jaskiw, S. Chrapusta, F. Karoum, D. R. Weinberger. Brain Res. 585, 1-6 (1992).

20. B. K. Lipska, G. E. Jaskiw, D. R. Weinberger. Neuropsychopharmacology 9, 67-75 (1993).

21. S. J. Clapcote, T. V. Lipina, J. K. Millar, S. Mackie, S. Christie. Neuron (2007); (in press)

22. H. Ogura, J. Aruga, K. Mikoshiba. Behav. Genet. 31, 317-24 (2001).

23. M. A. Geyer, K. L. McIlwain, R. Paylor. Mol. Psychiatry 7, 1039-53 (2002).

24. M. Fritzsche. Med. Hypotheses 56, 638-43 (2001).

25. H. Stefansson, E. Sigurdsson, V. Steinthorsdottir, S. Bjornsdottir, T. Sigmundsson et al. Am. J. Hum. Genet. 71, 877-92 (2002).

26. G. Torres, B. H. Hallas, V. A. Vernace, C. Jones, K. W. Gross et al. Brain. Res. Bull. 62, 315-26 (2004).

27. G. E. Duncan, S. S. Moy, A. Perez, D. M. Eddy, W. M. Zinzow et al. Behav. Brain. Res. 153, 507-19 (2004).

28. N. Hiroi, H. Zhu, M. Lee, B. Funke, M. Arai et al. Proc. Natl. Acad. Sci. U.S.A. 102, 19132-7 (2005).

29. S. L. Almond, R. L. Fradley, E. J. Armstrong, R. B. Heavens, A. R. Rutter et al. Mol. Cell .Neurosci. 32, 324-34 (2006).

30. C. Kellendonk, E. H. Simpson, H. J. Polan, G. Malleret, S. Vronskaya et al. Neuron 49, 603-15 (2006).

31. C. Mehler-Wex, P. Riederer, M. Gerlach. Neurotox Res. 10, 167-79 (2006).

32. E. K. Miller. Nat. Rev. Neurosci. 1, 59-65 (2000).

33. J. K. Seamans, C. R. Yang. Prog. Neurobiol. 74, 1-58 (2004).

34. C. Damgaard. J. Theor. Biol. 227, 197-203 (2004).

35. H. E. Wahlstrom, R. Akimoto, D. Endres, O. Kolterman, A. R. Moossa. Transplantation 53, 1190-5 (1992).

36. M. J. MacLellan, A. E. Patla. Exp. Brain. Res. 173, 521-30 (2006).

37. E. J. Calabrese. Crit. Rev. Toxicol. 31, 425-70 (2001).

38. J. I. Javaid. J. Clin. Pharmacol. 34, 286-95 (1994).

39. O. Agid, S. Kapur, T. Arenovich, R. B. Zipursky. Arch. Gen. Psychiatry 60, 1228-35 (2003).

40. O. Agid, P. Seeman, S. Kapur. J. Psychiatry Neurosci. 31, 93-100 (2006).

41. M. Li, P. J. Fletcher, S. Kapur. Neuropsychopharmacology 32, 263-72 (2007).

Share your thoughts

Leave a Reply

You must be logged in to post a comment.