Abstract
The role of serotonin in depression and antidepressant treatment remains unresolved despite decades of research. In this paper, we make three major claims. First, serotonin transmission is elevated in multiple depressive phenotypes, including melancholia, a subtype associated with sustained cognition. The primary challenge to this first claim is that the direct pharmacological effect of most symptom-reducing medications, such as the selective serotonin reuptake inhibitors (SSRIs), is to increase synaptic serotonin. The second claim, which is crucial to resolving this paradox, is that the serotonergic system evolved to regulate energy. By increasing extracellular serotonin, SSRIs disrupt energy homeostasis and often worsen symptoms during acute treatment. Our third claim is that symptom reduction is not achieved by the direct pharmacological properties of SSRIs, but by the brain's compensatory responses that attempt to restore energy homeostasis. These responses take several weeks to develop, which explains why SSRIs have a therapeutic delay. We demonstrate the utility of our claims by examining what happens in animal models of melancholia and during acute and chronic SSRI treatment.
Introduction
Depression is a heterogeneous suite of states characterized by sad mood and anhedonia (an inability to experience pleasure) (Hyman, 2010, Insel and Charney, 2003). Depressive states share some genes and neurobiology in common, but they otherwise differ in symptom and etiology (Akiskal and Akiskal, 2007, Dantzer et al., 2008, Flint and Kendler, 2014, Lux and Kendler, 2010, Maier and Watkins, 1998, Parker, 2000, Raison and Miller, 2013, Sullivan et al., 2012). For instance, depressive symptoms can occur in response to infection (called sickness behavior) or starvation (Hart, 1988, Keys et al., 1950), though the symptoms are not considered pathological in these contexts (Andrews and Durisko, in press, Dantzer, 2001, Engel and Schmale, 1972). In the fifth edition of the Diagnostic and Statistical Manual for Mental Disorders (DSM-5), the diagnostic category of major depression envelops some of the symptomatic heterogeneity by allowing for variability in weight, sleeping, and psychomotor activity (Table 1) (APA, 2013).
Episodes of major depression may be further subdivided into more precise phenotypes. Melancholia (weight loss, insomnia, and agitation/retardation) is considered by many to be the “biological core of depression” (Akiskal and Akiskal, 2007, p. 46). It is the most common and reliably diagnosed subtype, often accounting for 50% or more of clinical episodes (Angst et al., 2007, Taylor and Fink, 2008, Xiang et al., 2012). Melancholia is associated with heightened hypothalamic-pituitary-adrenal (HPA) activity (Taylor and Fink, 2008), which is a physiological indicator of stress (Chrousos, 2009). While it was formerly called endogenous depression, melancholia is in fact associated with stressful life events that are often serious or highly private in nature (Harkness and Monroe, 2002, Leff et al., 1970, Mundt et al., 2000, Willner et al., 1990). Atypical depression (weight gain, hypersomnia, and retardation) is the other major subtype, but it is heterogeneous and not well understood (Stewart et al., 2007).
Despite decades of research, the role serotonin plays in depressive phenotypes has not been conclusively determined. The original clue that monoamines (serotonin, norepinephrine, and dopamine) were involved in depression came from two serendipitous discoveries (Baumeister et al., 2003, Valenstein, 1998). First, during the investigations of iproniazid as a treatment for tuberculosis and imipramine as a treatment for schizophrenia, clinicians reported that these drugs could reduce depressive symptoms. An effort was then made to find a common pharmacological property that could explain their antidepressant effect. Eventually, researchers found that iproniazid inhibits the enzymes that breakdown the monoamines, while imipramine blocks the serotonin transporter (SERT) and the norepinephrine transporter (NET). Second, clinical observations suggested that reserpine, a drug known to deplete monoamines, increased depressive symptoms. These findings appeared to solve the puzzle. By preventing the breakdown of norepinephrine and serotonin, or preventing their clearance from the synapse, iproniazid and imipramine appeared to increase forebrain monoamine levels. The monoamine-enhancing effect of antidepressant medications (ADMs), coupled with the depression-inducing effects of reserpine, suggested that depression was caused by reduced monoamine neurotransmission (Everett and Toman, 1959, Jacobsen, 1964, Schildkraut, 1965).
Other researchers soon suggested that serotonin was the most important monoamine (Coppen, 1967). Often it is called the ‘monoamine hypothesis’ or the ‘serotonin hypothesis.’ We refer to it as the low serotonin hypothesis because it proposes a particular direction. Researchers then focused on the creation of drugs that could increase synaptic serotonin without perturbing other monoamines by selectively binding to the serotonin transporter (SERT). This research effort was successful, and the selective serotonin reuptake inhibitors (SSRIs) are now among the most widely prescribed medications (Olfson and Marcus, 2009, Olfson et al., 2002).
However, many problems with the low serotonin hypothesis have prompted a reassessment of serotonin's role in depression (see Box 1). Although the idea that a single neurochemical is the cause of depression is now considered simplistic, the low serotonin hypothesis still lies at the foundation of most research on depression (Albert et al., 2012). It is generally thought that reduced serotonin transmission is one of the more distal factors in the neurological chain of events that regulate depressive symptoms (Krishnan and Nestler, 2008). The fact that ketamine (which blocks a glutamate receptor) has rapid antidepressant effects lends support to the hypothesis that depressive symptoms are more proximally controlled by glutamate transmission in frontal regions (Mahar et al., 2014, Popoli et al., 2012). Others propose serotonin does not exert any regulatory control over depressive symptoms (Kirsch, 2010, Lacasse and Leo, 2005). Still others have suggested serotonin transmission is elevated in depression (Andrews and Thomson, 2009, Petty et al., 1994, Zangen et al., 1997).
In this paper, we make three major claims. The first claim, discussed in Section 2, is that serotonin transmission is elevated in multiple depressive phenotypes, including melancholia, infection, and starvation. We refer to this as the high serotonin hypothesis. What constitutes evidence of serotonin transmission is the key to the evaluation of this hypothesis. Since depression is a persistent state, reliable indices of stable serotonin transmission are particularly relevant. The 5-HIAA/5-HT ratio is the most reliable index of stable serotonin transmission, although 5-HIAA is also used (Shannon et al., 1986). While the literature on depressed patients is necessarily limited due to the methodological difficulties measuring serotonin and 5-HIAA in the human brain, the most pertinent studies support the high serotonin hypothesis. In non-human animal models of depression—where these indices can be measured more readily—abundant evidence supports the high serotonin hypothesis.
The primary challenge for the high serotonin hypothesis is explaining how ADMs, nearly all of which rapidly increase extracellular serotonin, reduce depressive symptoms. Our second claim, discussed in Section 3, is crucial to resolving this paradox. Specifically, we argue that the evolved function of the serotonergic system is energy regulation—which we define as the coordination of metabolic processes with the storage, mobilization, distribution, production and utilization of energetic resources to meet adaptive demands (Table 2).
In the brain and throughout the body, serotonin is homeostatically regulated (Best et al., 2010, Gershon and Tack, 2007, Mercado and Kilic, 2010). The energy regulation hypothesis predicts that the homeostatic equilibrium level of serotonin transmission is elevated in situations that require limited energetic resources to be reallocated among metabolically expensive processes: growth, reproduction, physical activity, maintenance, immune function, and cognition. Table 3 shows there is a stable increase in serotonin transmission to the hypothalamus in both positive and negative mood states where energy must be reallocated for prolonged periods of time. Similarly, the effects of SSRIs are state-dependent. Depending on the context, SSRIs can increase or decrease anxiety (Robert et al., 2011), motor activity (Altemus et al., 1996, Page et al., 1999), anhedonia (Branchi et al., 2013, Harrison et al., 2001), and neurotrophin signaling (Bland et al., 2007, Freitas et al., 2013, Hellweg et al., 2007, Rasmusson et al., 2002, Van Hoomissen et al., 2003). Thus, serotonin cannot be simply described as an ‘upper’ or a ‘downer’; its symptomatic effects depend on the organism's state (i.e., whether it is infected, starving, satiated, physically exhausted, sexually exhausted, etc.).
Table 4 lists the symptoms of three reliably diagnosed depressive states: sickness behavior, starvation depression, and melancholia. Each involves an altered balance between metabolically expensive processes (Fig. 1). In sickness behavior, limited energetic resources are devoted to immune function at the expense of growth and reproduction. In starvation depression, energy is devoted to maintenance functions at the expense of growth, reproduction, and immune function. In melancholia, there is an upregulation in sustained cognition at the expense of growth and reproduction. The energy regulation hypothesis suggests serotonin transmission is elevated in these states to coordinate tradeoffs in energy allocation. In melancholia, this tradeoff is coordinated by serotonin transmission to various regions, including the hypothalamus, amygdala, hippocampus and lateral prefrontal cortex (PFC) (Fig. 2). In the hippocampus and lateral PFC, the processes involved in sustained cognition are energetically expensive and can only be sustained with aerobic glycolysis (the generation of lactate from the metabolism of glucose stored in astrocytes).
Our third major claim, discussed in Section 4, is that the direct pharmacological effects of SSRIs are not responsible for symptom reduction. Rather, by rapidly increasing extracellular serotonin, SSRIs cause a disruption in energy homeostasis (the state-dependent balance between energetically expensive processes that existed prior to medication), and a worsening of symptoms. For instance, in melancholia, serotonin transmission to the PFC causes an increase in energetically expensive glutamatergic activity (Fig. 3B), which is exacerbated during acute SSRI treatment (Fig. 3C). We argue that symptom reduction is due to the compensatory changes made by the brain's homeostatic mechanisms that attempt to restore energy homeostasis (Fig. 3D). These compensatory changes take several weeks to develop, which explains why symptoms fail to alleviate for several weeks after the initiation of SSRI treatment (the therapeutic delay).
In Section 5, we show how these claims help explain what is happening in non-human animal models of melancholia and during acute and chronic treatment with SSRIs. We conclude with implications and suggestions for future research.
Section snippets
Serotonin is elevated in multiple depressive phenotypes
It is currently impossible to measure 5-HT in the living human brain because it requires invasive techniques (Marsden, 2010). Moreover, serotonin cannot cross the blood brain barrier (Bouchard, 1972, Genot et al., 1981), so peripheral measures do not accurately reflect brain levels.
Some studies use molecular in vivo neuroimaging techniques to attempt to infer changes in endogenous serotonin levels (Bhagwagar et al., 2007, Savitz et al., 2009, Stockmeier, 2003). These techniques can measure.
The energy regulation function of the serotonergic system
In this section of the paper, we propose a novel hypothesis for the evolved function of the serotonergic system, which includes serotonin, its receptors, SERT, and other components that help regulate serotonin or its effects. Our hypothesis owes much to the research of Efrain Azmitia on the evolution of serotonin (Azmitia, 2001, Azmitia, 2007, Azmitia, 2010). One of our novel contributions is to explicitly identify the evolution of the mitochondrion as the likely point on the tree of life where
The homeostatic response to SSRIs and symptom reduction
In this section, we argue that depressive symptoms are reduced over several weeks of SSRI treatment, not by their direct pharmacological properties, but due to the compensatory responses of the brain attempting to restore energy homeostasis.
What is serotonin doing in melancholia?
Since the effects of serotonin are state-dependent, we demonstrate the utility of our hypotheses in explaining what happens in the melancholic state. In melancholia, the symptoms reflect a trade-off in which energy is reallocated toward cognition at the expense of growth and reproduction. We suggest that the elevation in serotonin transmission coordinates this trade-off and helps explain many of the symptoms of melancholia.
Conclusion and future directions
The reigning paradigm conceptualizes depression as a state of reduced serotonin transmission. In this paper we have reviewed a large body of evidence indicating that the opposite appears to be true. For the depressive phenotypes we have considered—sickness behavior, starvation depression, and melancholia—serotonin transmission to multiple brain regions appears to be elevated. Others have suggested serotonin transmission is elevated in depression (Andrews and Thomson, 2009, Petty et al., 1994.
Acknowledgments
The authors gratefully acknowledge the following people for comments, conversations or correspondence that influenced the manuscript: Marie Banich, Dick Day, Denys DeCatanzaro, Zac Durisko, Steve Gangestad, Ed Hagen, Chris Lowry, Steve Maier, and Onkar Marway. Three anonymous reviewers provided thoughtful comments that greatly improved the manuscript.
Comments