Mechanisms of Illness Progression in the Recurrent Affective Disorders
Along with genetic vulnerability, multiple environmental factors convey liability to illness progression, including: (1) distal and proximal stressors; (2) recurrence of episodes; and (3) comorbid cocaine abuse. Recurrence of each of these can increase responsivity (sensitize) to themselves and cross-sensitize to the two other factors and drive illness progression as seen clinically in increases in cycle acceleration, severity or duration of episodes, treatment refractoriness, disability, cognitive dysfunction, and premature death. Some mechanisms appear common to all three types of sensitization, such as decreases of brain-derived neuroprotective factor (BDNF) in hippocampus and blood, as well as increases in BDNF in the nucleus accumbens, suggesting the possibility that single treatments could ameliorate several of these factors at once. A potential example is N-acetylcysteine (NAC), which decreases bipolar affective illness severity (Berk et al. Biol Psychiatry 64:468–475, 2008) and cocaine reinstatement and craving (Baker et al. Ann N Y Acad Sci 1003:349–351, 2003; LaRowe et al. Am J Addict 15:105–110, 2006). Mechanisms of illness progression also involve epigenetic changes and add further rationale to the existing empirical clinical evidence of the importance of early recognition, treatment, and prevention of affective episodes. Adequate treatment could prevent or ameliorate both the increases in pathological factors and erosion of adaptive factors that propel illness exacerbation and treatment resistance. This view of the sensitization and cross-sensitization among stressors, episodes, and abused substances should lead to a fundamental re-conceptualization of the recurrent affective disorders not as benign, isolated episodes of “mental” illness, but as severe, potentially progressive and lethal medical disorders of brain and body that deserve careful life-long monitoring and treatment.
KeywordsKindling Sensitization Depression Bipolar disorder Brain-derived neurotrophic factor (BDNF) Stress Epigenetics Cocaine
One hundred years ago, Emil Kraepelin recognized the fundamental recurrent and progressive components of bipolar affective disorder, which he separated from schizophrenia based on the former’s episodic nature, pleomorphic affective states, and usual return to baseline functioning between episodes. In contrast, schizophrenia appeared to be a more chronic, progressive illness, with a less variant downhill course associated with cognitive deterioration or dementia praecox (Kraepelin 1921).
Kraepelin was a keen observer of bipolar illness and charted episodes of mania, depression, and mixed states in individual patients. He remarked on the extreme pleomorphic nature of the illness and its unpredictability. However, he also observed a general overall tendency for episodes to emerge with a shorter well interval as a function of the number of prior episodes (cycle acceleration). At the same time, he saw increasing illness autonomy and a progression to more spontaneous episodes; i.e., he observed that initial episodes often appeared to be triggered by psychosocial stressors, but with sufficient repetition, episodes began to emerge spontaneously. He also suggested that there was a genetic component of illness onset and progression, as well as an interaction with the environment in the form of episode acceleration (which we have labeled episode sensitization) and increasing sensitivity to stressors to the point that actual stressors are no longer required to precipitate episodes; anticipated or imagined stressors might be sufficient (conditioned or context-dependent stress sensitization).
These fundamental progressive aspects of the illness have now been recognized in both unipolar and bipolar affective disorders and replicated in multiple longitudinal studies (Post 1992; Post and Post 2004; Kessing and Andersen 2005; Post and Leverich 2008; Post and Miklowitz 2010). In the several instances where they have not been replicated, short-term observations, small numbers of individuals observed, misinterpretations of the stress-sensitization concept, and other methodological shortcomings appear to be pertinent potential explanatory variables (Post 2004; Kessing and Andersen 2005; Dienes et al. 2006).
Among the strongest evidence for increased vulnerability to relapse as a function of number of prior episodes are the data of Kessing et al. (1998; Kessing and Andersen 2005) from the Danish Case Registry of more than 20,000 individuals. In this population, the incidence and latency to a depressive relapse (for either unipolar or bipolar depressive hospitalization) were related to the number of prior hospitalizations for depression. Kessing (2008) also confirmed earlier clinical observations that severity of depressive episodes also increases over the course of illness, findings consistent with our large series of graphic portrayals of individual’s illness course and response to treatment (Post and Leverich 2008).
Kraepelin’s other essential observation was the progression of episodes from those that are triggered by psychosocial stressors to those that could occur more spontaneously as a function of the number of prior episodes. This has been most elegantly demonstrated by Ken Kendler et al. (2000, 2001) in patients with unipolar depression. These investigators found that psychosocial stressors were most strongly implicated in the triggering of the first seven episodes, but with further occurrences, stressors appeared less crucial to episode recurrence. These data are consistent with a stress-sensitization phenomenon, i.e., episodes are increasingly more readily triggered to the point that minimal stressors (or, apparently, none at all) are sufficient to precipitate episodes. Thus, the fact that episodes later in the course of the illness can also be triggered by psychosocial stresses does not contradict the concept of stress sensitization and the kindling-like progression of recurrent affective episodes.
The Kindling and Sensitization Models for Illness Progression
Given the fundamental clinical observations noted above (Post 1992; Kendler et al. 2001; Kessing and Andersen 2005; Post and Leverich 2008; Post and Miklowitz 2010), for progressive changes in frequency, severity, and spontaneity of affective episodes, we examined two different animal models—behavioral sensitization to psychomotor stimulants and electrophysiological kindling—that might reflect aspects of these progressive changes (Post 1992, 2007a, 2008).
While studying the potential antidepressant effects of cocaine in refractory depression, we became aware of a considerable literature indicating that repetition of cocaine administration in animals might lead to an increase in responsivity in several domains (Post et al. 1987). One was behavioral sensitization to cocaine, in which an increased hyperactive and stereotypic response would be observed upon repeated drug administration, rather than a tolerance process. In instances of repeated administration of higher but stable doses of cocaine, the development of cocaine-induced seizures could occur to a dose that was previously sub-convulsant.
This latter progression appeared to resemble the course of electrical kindling, in which repeated electrical stimulation of the amygdala on a daily basis for 1 s came to evoke increasing duration, complexity, and anatomical spread of amygdala after discharges, which correlated with progressive behavioral changes, culminating in the occurrence of full-blown bilateral tonic-clonic seizures to a stimulation which was previously ineffective (Goddard et al. 1969). With repeated stimulation and afterdischarge spread, there was a concomitant progression from behavioral arrest to unilateral and then bilateral forepaw involvement, to a full-blown major motor seizure with rearing and falling (Racine 1972). Correlated with these physiological and behavioral changes were a spatio-temporal spread of induction of transcriptions such as c-fos, neuropeptides such as TRH, and neurotrophic factors such as brain-derived neuroprotective factor (BDNF). Following a sufficient number of these stimulation-induced seizures, spontaneous seizures (in the absence of exogenous amygdala stimulation) would emerge.
The same type of seizure progression appeared to be induced with repeated high intraperitoneal doses of cocaine, as well as lidocaine which had equi-potent local-anesthetic effects but lacked cocaine’s psychomotor stimulant properties (Post et al. 1975, 1987). With once-daily local-anesthetic administration at doses that initially were non-convulsive, seizures began to occur intermittently to regularly following each injection. Particularly with lidocaine (which was well tolerated and with which animals could survive runs of seizures lasting as long as 2 h), the daily occurrence of enough of these drug-induced seizures would result in the appearance of seizures even when no lidocaine was given. This time course of progressive seizure susceptibility paralleled that of amygdala-kindled seizures, and in light of the local anesthetics inducing amygdala activation and local afterdischarges, we labeled this process ‘pharmacological kindling’ (Post et al. 1975, 1987).
In contrast to the stimulation of the amygdala by electrical or pharmacological means, the behavioral sensitization to cocaine appeared to more closely mirror sensitization to other psychomotor stimulants such as amphetamine or methylphenidate that did not cause seizures. Instead, hyperactivity became more severe and stereotypies more frantic and constricted. Some types of intermittent stressors that showed sensitization, such as tail pinch or mild foot shock, would show cross-sensitization to the stimulant-induced behavioral sensitization and vice versa (Antelman 1988; Kalivas and Stewart 1991). In particular, animals stressed early in life appeared more likely to respond to and adopt cocaine and amphetamine self-administration compared with litter-mate controls (Post and Post 2004). Conversely, stimulant-sensitized animals were more stress-responsive as well.
While stimulant-induced behavioral sensitization could be induced by merely repeating the same dose of the drug, it became apparent that sensitization was stronger when the environmental context cues remained the same, and the drug was repeatedly administered in the same environment (Post et al. 1987). These studies revealed a strong context dependency or conditioned component of cocaine sensitization. When cocaine was administered in one environment and the animals were tested in another, there might be a complete absence of the sensitized motor or stereotypic response (Post et al. 1987; Post 2007a).
Thus, we saw that cocaine-induced behavioral sensitization and kindling (either electrical or pharmacological) represented two very different models of illness progression, i.e., the measured endpoints were very different – behavioral hyperactivity and stereotypy in one instance and seizures in another. The neurochemistry and the anatomy involved differed as well. Behavioral sensitization initially involved dopaminergic mechanisms in the VTA/nucleus accumbens pathway, and later other more widespread areas (Kalivas and Stewart 1991; Kalivas and Volkow 2005; Kalivas 2008; Kalivas and O’Brien 2008), while kindled seizures involved the induction of local afterdischarges and changes in gene expression that spread progressively from the amygdala to increasingly more distant areas of brain (Clark et al. 1991).
Clinical parallels in the course of affective illness with kindling and sensitization
Strength of observations in
Phenomena modeled by:
Increase vulnerability to relapse as function of number of prior episodes (cycle acceleration/shorter well intervals)
External triggers common in initial episodes, less necessary with later episodes
Episodes may increase in frequency, severity, duration, complexity
Conditioned cues play a major role in behavioral manifestations
Different drugs are effective in different stages of illness progression
Tolerance may develop to a previously effective prophylactic treatment
Episode recurrence and cyclicity may reflect the ratio of pathological to adaptive factors
Yet, some aspects of the clinical phenomenology of cocaine abuse behaviors in humans do closely mirror hypomanic and manic illness progression in the primary bipolar disorders. That is, initial use of cocaine is typically associated with euphoria, increased energy, and a sense of well-being, but with sufficient repetition of cocaine use increasing amounts of anxiety, dysphoria, context-dependent paranoia, full-blown panic attacks, and even the emergence of a paranoid psychosis are observed (Post et al. 1987). A similar progression from euphoric hypomania to dysphoric mania and paranoid psychosis may also occur in primary bipolar disorder. Thus, components of cocaine and amphetamine-induced behavioral sensitization in animals mirror many of the progressive aspects of stimulant abuse and addiction in humans, which in turn parallel changes that can occur in the course of repeated manic episodes. Moreover, in bipolar patients with a history of stimulant administration, there is an increased incidence of dysphoric mania compared to those without such a history (Kalivas and Volkow 2005; Post 2007a). This suggests the cross-sensitization from the stimulant-induced behavioral progression to that occurring naturalistically in the longitudinal course of bipolar disorder.
As the differential phenomenology, neuroanatomy, neurochemistry, and pharmacology of the sensitization and kindling models were revealed, one could ask whether some of the components and mechanisms of either type of progressive syndrome were pertinent to the longitudinal course and treatment of the affective disorders. Ultimately, the sensitization and kindling models would have heuristic value even as non-homologous models of affective illness in suggesting potential principles of illness evolution that could then be directly tested in the clinic.
For example, as noted above, studies with c-fos and messenger RNA for a variety of neuropeptides and neurotrophic factors revealed that there was a spatial–temporal spread of changes in gene expression that accompanied kindled seizure evolution (Clark et al. 1991; Rosen et al. 1992). Increasingly wider and diverse neuroanatomical and biochemical substrates appeared to be evolved as a function of the stage of kindled seizure evolution, from (A) the initial or developmental stage; to (B) the middle phase of consistent trigger of full-blown seizure; to (C) the late stage of spontaneity. Not surprisingly, pharmacotherapeutic interventions also were found to differ as a function of stage of kindled seizure evolution; some agents that were effective in (A) the initial developmental stages of kindling were not effective against (B) full-blown seizures, and even more often drugs showed the opposite pattern (Post 2008). Moreover, some drugs such as diazepam that were highly potent in preventing (A) development and (B) full-blown seizures were no longer effective against (C) the spontaneous variety. In a double dissociation, phenytoin showed the opposite pattern—it was a weak or ineffective agent in the (A) and (B) stages of triggered kindling seizure evolution, but was potent against (C) the late, spontaneous seizures (Pinel 1983).
One could then ask whether some of these same general principles of differential biochemistry, neuroanatomy, and pharmacology as a function of kindled seizure stage evolution might be applicable to illness evolution and treatment in the affective disorders, keeping in mind the caveat that the specific mechanisms underlying affective episodes and the drugs for their treatment may be different from those involved in sensitization and kindling. Nonetheless, the general concept that the neurobiology and effective pharmacology may differ as a function of stage of syndrome evolution can be examined for its indirect predictive validity in the affective disorders.
The kindling model also helped clarify the concept that some seizure-induced neuro-chemical alterations were related to the primary pathological process of kindling and its memory trace, while others had putative adaptive or anticonvulsant properties (Post and Weiss 1992; Post 2007a, 2008). This idea was derived from the observation that the anticonvulsant effects of some drugs, such as carbamazepine or lamotrigine, against full-blown amygdala-kindled seizures depended on the immediately prior occurrence of kindled seizures and their presumptive seizure-induced biochemical alterations. In contrast, if the animal was put on vacation and no seizures were induced for 5–7 days (a timeframe associated with the loss of many seizure-induced adaptations such as increases in thyrotropin releasing hormone), anticonvulsants which were fully effective at a given dose would lose their anticonvulsant properties (Weiss et al. 1995).
It appeared that it was the ratio of pathological to adaptive changes induced by amygdala stimulation (as supplemented by exogenous medications) that influenced whether a seizure occurred or not. We then hypothesized that in a similar fashion, the ratio of pathological to adaptive factors (as supplemented by medications) would also account for whether an affective episode occurred or not (Post and Weiss 1996; Post 2007a, 2008). We suggested that similar to what happened following kindled seizures, episode-related increases in thyrotropin releasing hormone (TRH) may represent one of many adaptive endogenous antidepressive mechanisms, while the increases in corticotrophin-releasing hormone (CRH) and its downstream increases in cortisol may represent one of many pathological factors associated with episode occurrence and illness progression (Post and Weiss 1998). If this hypothesis proved valid, it would clearly have therapeutic implications. New therapeutic approaches to affective disorders would then be differentially targeted to either enhance and supplement the adaptive alterations or to inhibit or suppress the primary pathological ones (Post and Weiss 1998; Post 2007a).
Contingent Tolerance to the Anticonvulsant Effects of Drugs in Preventing Amygdala-Kindled Seizures: A Secondary Manifestation of Illness Progression Breaking Through Effective Pharmaco Prophylaxis
Similar to the initial progressive emergence of seizures upon repeated amygdala stimulation in the absence of treatment, we observed that after a period of excellent and full anticonvulsant response to a drug such as carbamazepine, amygdala-kindled seizures would begin to progressively break through previously effective doses of the drug with increasing frequency and severity, in an apparent tolerance mechanism. This type of tolerance development had a conditional or contingent component, because animals only became tolerant when the drug was administered and was in the brain prior to the amygdala stimulation. In contrast, if animals were repeatedly administered the same doses of carbamazepine on a once-daily basis, but immediately after a kindled seizure had occurred, they did not become tolerant (Weiss et al. 1995; Post 2008). We found that these breakthrough seizures in tolerant animals were associated with a loss of some of the adaptive putatively anticonvulsant changes in gene expression, such as increases in TRH or the alpha-4 subunit of the GABAA receptor. We partially validated the finding that the loss of seizure-induced TRH contributed to the tolerance process by infusing TRH bilaterally into the hippocampus in animals that were tolerant to the anticonvulsant effects of carbamazepine and found that there was renewed anticonvulsant efficacy.
We also found that if animals that were tolerant to the anticonvulsant effects of carbamazepine were then given a series of amygdala-kindled seizures while drug free, the occurrence of seizures in the absence of the drug was sufficient to reverse the tolerance process, and anticonvulsant efficacy reappeared. We postulated that these new seizures in the absence of the drug were associated with the renewed induction of putative seizure-induced anticonvulsant adaptations, such as the increases in TRH and GABAA receptors, which were now sufficient to enable the anticonvulsant effects to appear again.
These sets of observations also led us to hypothesize that some of the manipulations that would slow, prevent, or reverse tolerance development in the amygdala-kindling model of anticonvulsant tolerance might be pertinent to that observed in the affective disorders. Chronic long-term prophylaxis with antidepressants for recurrent unipolar depression is associated with sustained responses in the majority of patients. However, in a minority of individuals, there appears to be a tolerance process wherein depressions begin to break through a previously effective prophylactic treatment. The same has been observed for most of the mood stabilizers used in the long-term prophylaxis of bipolar disorder, including lithium, carbamazepine, lamotrigine, and to a lesser extent, valproate. One could examine whether some of the manipulations found to slow or reverse tolerance development in the kindled seizure model might be applicable to the affective disorders, and these possibilities could be directly tested in the clinic.
As we observed in the preclinical model, one could ask whether clinical tolerance would be slowed following: (1) earlier institution of long-term prophylaxis (before the accumulation of multiple episodes), (2) treatment using higher doses, (3) more efficacious drugs, and (4) drugs used in combination (Post 2007a, b). Again, the caveat must be emphasized that the specific drugs examined in the kindled seizure model may or may not be pertinent to prophylaxis of the affective disorders. A small subgroup of patients experience tolerance during the long-term administration of lithium for bipolar illness, and some of the same principles noted above for slowing tolerance to anticonvulsant drugs might nonetheless be applicable to lithium even though it is not effective against amygdala-kindled seizures.
Neurobiological Abnormalities in the Recurrent Affective Disorders as a Function of Number of Episodes or Duration of Illness
Many of these decreases in frontal cortical structure or function are cross-validated with decrements in neuronal and glial markers such as N-acetyl aspartate (NAA) and glial fibrillary acidic protein (GFAP), respectively. In addition, there is evidence on a micro-anatomical level from autopsy studies for deficits in neuronal and glial size or number, consistent with increased loss or inadequate production of cellular elements in those areas of brain. There are also major alterations in sleep and other physiological and somatic systems, and in addition, 50% or more patients with either unipolar or bipolar depression show evidence of hypercortisolemia, often accompanied by deficits in somatostatin and increases in CRF measured in cerebrospinal fluid (CSF). There is also evidence for decreased reactivity of growth hormone to a variety of challenges and decreased TSH stimulation by TRH, potentially reflecting TRH increases down-regulating TRH receptors.
Cocaine addiction is also associated with frontal hyperactivity (again in proportion to the severity of depression observed during abstinence), as well as increased activity of the amygdala and nucleus accumbens (Kalivas and Volkow 2005), suggesting that many of cocaine’s effects may overlap with those of the primary affective disorders and potentially exacerbate them.
Given that genetic vulnerability interacts with environmental experience in the form of stressors and episodes, some of the neurobiological abnormalities identified in affective illness might be constant based on genetic inheritance, while others could potentially be cumulative, based on the accretion of stressors and the accumulation of recurrent affective episodes. Moreover, as there is a high comorbidity of substance abuse in the recurrent unipolar and bipolar disorders (Sonne et al. 1994; McElroy et al. 2001; Wilens et al. 2004), alcohol and substance abuse generate a third potential environmental impact on the neurobiology of affective illness that could be associated with illness progression.
To the extent that there are stress- and episode-sensitization and kindling-like components to affective illness progression, some evidence of these processes should be observable, for example, if neurobiological abnormalities were related to the number of prior episodes or to the duration of illness, it would at least raise the possibility that they were the result of episode occurrence and a reflection of illness progression. In fact, a substantial number of the reported abnormalities in the unipolar and bipolar affective disorders are greater in those with more prior episodes or longer duration of illness.
Neurobiological correlates of number of episodes or duration of illness in the recurrent affective disorders
UP D /or BP
# of episodes
First Author et al. (year)a
Subgenual metabolism decrease
Anterior cingulate vol. decrease
Ventral prefrontal vol. decrease
Decrease in volume
Synaptic markers (autopsy)
Decrease in gray matter
Temporal lobe gray matter decrease
# m, d
Moorhead (2007a, b)
Robinson and Ferrier (2006)
Lateral ventricles increase
# m, d
Cognitive dysfunction, disability
# m, d
Dementia in late life risk doubled
Kessing and Anderson (2004)
In adults: volume increase
In children: volume decrease
Increase dexamethasone escape
Increase dexamethasone/Corticotropin releasing hormone; risk relapse
Increase plasma cortisol
Decrease in serum BDNF
Allostatic load increased
Coronary heart disease and increased blood pressure
Perhaps the most clear-cut data on this topic of illness progression come from the Danish Case Registry (Kessing and Andersen 2004), who demonstrated that the occurrence of two major episodes of either unipolar or bipolar depression was associated with an incidence of late life dementia equal to that of the general population, but if a patient experienced four episodes in total, this was associated with a doubling of the risk for the occurrence of dementia later in life, and each depressive episode thereafter increased the likelihood further. Conversely, lithium treatment either by its episode preventing or neurotropic/neuroprotective properties (as discussed below) is associated with a reduced incidence of dementia (Kessing et al. 2008), further suggesting that adequate treatment intervention could prevent this type of illness progression.
Given the wealth of evidence summarized in Table 2, further studies are warranted to examine in a more systematic and prospective fashion whether episode prevention could mitigate some of these abnormalities and the causal mechanisms inferred from them. However, this process should proceed with the caveat that it is critical to differentiate between abnormalities that are pathological versus adaptive, as inhibiting the latter could prove counterproductive.
BDNF as a Putative Mediator of Illness Progression
Brain-derived neuroprotective factor is affected by stressors and substances of abuse (such as cocaine), and BDNF is decreased in blood in association with each affective episode (both manic and depressive). Thus, examining the potential role of BDNF in illness progression may give some hints as to the mechanisms involved (Duman and Monteggia 2006; Post 2007b).
The data are clear that repeated neonatal stress results in decrements in hippocampal BDNF, and in some instances this persists into adulthood and may also include decreased BDNF in the prefrontal cortex (Roceri et al. 2004; Roth et al. 2009). Accompanying these changes are a decrease in the set point for adult neurogenesis (Coe et al. 2003), potentially yielding vulnerability to a variety of illnesses based on lesser numbers of cells produced and their decreased survival because of low BDNF (Gomez-Pinilla and Vaynman 2005; Champagne and Meaney 2006).
Some parallel changes have been observed in humans. In a study of bipolar patients, those with a history of early physical or sexual abuse had significantly lower levels of BDNF than patients without such a history (Kauer-Sant’Anna et al. 2007). While it is not entirely clear what all the origins of serum BDNF are, considerable evidence suggests that there is a close relationship between peripheral and CNS levels. If BDNF is administered directly into specific brain areas, its effects are also positive in animal models of depression. In concert with the observations that all antidepressant modalities increase BDNF, these data suggest that BDNF could be one element in the final common pathway of antidepressants from a variety of different chemical classes (Duman and Monteggia 2006; Martinowich et al. 2007; Post 2007b).
Repeated stressors in adult rodents also decrease hippocampal BDNF, and this has been shown to be directly linked to the manifestation of depressive behaviors in the model of defeat stress (Berton et al. 2006; Tsankova et al. 2006, 2007). In this model, an intruder rodent is placed in the home cage of another larger rodent and is thus subject to violent attack, such that the animal would be killed if it were not protected by a glass barrier. The intruder animal is repeatedly introduced into the home cage territory of the resident rat, which it can see and smell. This repeated stressor results in the manifestation of depressive-like behavior and decrements in BDNF in the hippocampus of the intruder rat. However, if the animals are treated with antidepressants or if BDNF in the hippocampus is protected with genetic manipulations that increase its production, the chemical and behavioral changes are prevented (Tsankova et al. 2006).
At the same time, defeat stress results in increases in BDNF in the dopaminergic pathway from the midbrain ventral tegmental area (VTA) to the nucleus accumbens mediating locomotor activity, motivation and hedonic reward. Like the BDNF decreases in the hippocampus, if the BDNF increases are prevented in the nucleus accumbens, defeat stress behavior does not occur (Berton et al. 2006). Thus, it appears that opposite alterations in BDNF in the glutamatergic hippocampal system and in the dopaminergic nucleus accumbens system are both critically linked to the emergence of defeat stress behaviors. In humans who died of suicide, BDNF and its Trk B receptor are likewise decreased in hippocampus, and BDNF in the nucleus accumbens is also increased (Krishnan et al. 2008).
It is striking that repeated cocaine administration, which produces behavioral sensitization, is also associated with increases of BDNF in the nucleus accumbens, and like the defeat stress scenario, if the BDNF increases are prevented, sensitization does not occur (Nestler and Carlezon 2006). Moreover, in some instances, cocaine administration is associated with decrements in BDNF in the hippocampus, paralleling the changes induced by defeat stress. These opposite changes in BDNF in two neural substrates, each of which is sufficient for generating repeated defeat stress behaviors and cocaine-induced behavioral sensitization, could in part mediate the cross-sensitization observed between these two very different types of environmental stimuli. Another model of manic-like behavior induced by disruption of the circadian rhythm CLOCK gene is associated with increases in reward value for cocaine, concomitant with increases in dopaminergic activity and burst firing in the VTA (Roybal et al. 2007).
It is not clear why increases in BDNF in the nucleus accumbens should be associated with both defeat stress and cocaine sensitization, but it is possible that this mechanism is involved in the habit memory system encoded in the striatum (Mishkin and Appenzeller 1987). In contrast to conscious representational memory, which is mediated by amygdalar and hippocampal substrates, the habit memory system is thought to be unconscious and related to learning based on repetition.
It is particularly difficult to extinguish learned associations in the habit system. Animals and patients addicted to cocaine can have their cocaine reward cues desensitized for long periods of time and then show marked reinstatement behavior upon re-exposure to either cues signaling the availability of cocaine, the environment in which cocaine was available, or stressors (Kalivas and Volkow 2005). Similarly, patients who believe that they have learned to adequately cope with stressors that initially were observed to induce their affective episodes often find that new episodes are precipitated by minimal stressors or even none at all. These parallel observations of cocaine reinstatement and illness recurrence also raise the possibility that the stressor- and cocaine-induced cross-sensitization could be one of the reasons for the high comorbidity of substance abuse and recurrent affective disorders (Sonne et al. 1994; McElroy et al. 2001). Each may increase vulnerability to the other and potentially exacerbate the neurobiological abnormalities of the other as well. At the same time, such potential convergent mechanisms raise the possibility that a common therapeutic maneuver might help prevent both cocaine-induced reinstatement behavior and the vulnerability to recurrence of affective episodes.
The pioneering work of Peter Kalivas in his laboratory at the University of South Carolina demonstrated that NAC administration to animals could prevent cocaine reinstatement behavior (Baker et al. 2003; LaRowe et al. 2006; Zhou and Kalivas 2008). These observations were taken to the clinic, and it was found that NAC also decreased cocaine cravings in substance abusers. Kalivas had observed that cocaine reinstatement was associated with a marked increase in glutamate in the nucleus accumbens and reasoned that using NAC, which acted at the cystine/glutamate exchanger in that area would, via activation of a metabotropic glutamate receptors, prevent the exaggerated increases in glutamate associated with cues leading to cocaine reinstatement behavior. This is, in fact, what he observed.
At about the same time, Mike Berk from Australia began using NAC in refractory bipolar patients who were inadequately stabilized on their current psychotropic regimens (Berk et al. 2008). After administration of the compound for 3 or 6 months, NAC compared with placebo produced improvement on most clinical indices and particular improvement in depression. Berk and colleagues’ rationale for using NAC was that it increased glutathione biosynthesis and thus an anti-oxidant. It is not yet clear which of these different actions of NAC mediates its effects in bipolar patients.
To the extent that NAC works by similar mechanisms in decreasing cocaine reinstatement behavior and recurrent affective behavior, it would be an example of a possible single therapeutic agent able to modulate both of these processes, which have at least some neural substrates and neurobiology in common. Similarly, antidepressants not only block the development of defeat stress-induced depressive behaviors, but they also prevent a variety of stressors from decreasing BDNF in the hippocampus (Duman and Monteggia 2006; Post 2007b), providing another link between the sensitization to affective episodes and to stressors.
Earlier we noted that both depressive and manic episodes are associated with decrements in BDNF, often in proportion to the severity of the episode (Kauer-Sant’Anna et al. 2007; Post 2007b). Kauer-Sant’Anna et al. (2009) have shown that patients later in the course of their illness have greater decrements in BDNF compared with those earlier in the illness, suggesting a possible cumulative deficit in BDNF after multiple episodes. However, most studies indicate that BDNF will normalize with appropriate therapeutics and the achievement of a euthymic mood interval in between episodes. Thus, there may be both state- and trait-dependent alterations in BDNF that are associated with repeated affective episodes.
Mechanisms of illness progression in recurrent affective disorders
Increases in pathological processes
I. Passive loss of cellular resilience
A. Cellular endangerment and loss via increases in:
3. Oxidative stress
C. Risk factors
4. Sleep loss
7. Metabolic syndrome
A and B interact with C and result in cognitive decline, late life dementia, and loss of life expectancy (10–20 years)
B. Increased allostatic load and medical comorbidities
II. Active changes in gene expression and epigenetic process related to:
A. Episode accumulation
C. Substance abuse
D. Each sensitizes & cross-sensitizes to the others via:
1. Increases BDNF in nucleus accumbens
2. Decreases BDNF in hippocampus
E. Alterations by:
Loss of adaptive, protective factors
I. Neurobiological alterations
A. Loss of:
1. Neuroprotective factors (BDNF)
2. Antioxidants (glutathione)
3. Anti-inflammatory cytokines
4. Glucocorticoid receptors
6. Growth hormone response to clonidine
7. TRH response to TSH and Sleep
B. Epigenetic changes
1. Decreases in histone acetylation and induction of histone deacetylases
2. DNA methylation and silencing
II. Psychosocial processes via episodes, stressors, abused substances
1. Social support
2. Educational opportunities
3. Employment status
4. Health care insurance
5. Psychiatric and medical care access
6. Legal difficulties
7. Family and housing
These might include increases in the ratio of inflammatory to anti-inflammatory cytokines, increases in free radicals and other downstream consequences of increased oxidative stress (Ng et al. 2008; Andreazza et al. 2009), and deficits measurable at the level of mitochondrial and endoplasmic reticulum function (Andreazza et al. 2008; Kapczinski et al. 2008a). It is also noteworthy that both repeated cocaine and repeated stressors increase the accumulation of the transcription factor ∆ fos B, providing another overlapping mechanism potentially related to sensitization (Kumar et al. 2005; Levine et al. 2005; Nestler 2005; Tsankova et al. 2007).
Cross-Sensitization Among Stressors, Affective Episodes, and Substances of Abuse
It is possible that the therapeutic effects of some of the long-term treatments of the recurrent affective disorders may by mediated in part by effects on BDNF (Post 2007b; Post and Miklowitz 2010). The antidepressants used in the prophylactic treatment of unipolar depression increase BDNF and prevent stress from decreasing it in the hippocampus. The mood stabilizers used in the prevention of bipolar disorder, lithium and valproic acid, both increase BDNF and neurogenesis. Of the other agents used in bipolar patients, carbamazepine also increases BDNF, and the atypical antipsychotic quetiapine increases BDNF and prevents stress from decreasing it (Post 2007b). The atypical ziprasidone does not increase BDNF on its own, but prevents stress from decreasing it in the hippocampus. Interestingly, when the older typical antipsychotic (neuroleptic) haloperidol was used as a comparator, it not only did not prevent the stress-induced decreases in BDNF in the hippocampus, but actually exacerbated them (Park et al. 2009).
There is substantial clinical evidence that lithium induces clinically relevant neurotrophic and neuroprotective effects in patients with bipolar disorder, as revealed by increases in NAA observed via magnetic resonance spectroscopy (MRS) and increases cortical gray matter observable in MRI scans (Moore et al. 2000; Bearden et al. 2007; Martinowich et al. 2007). Sheline et al. (2003) found that greater compared with less use of antidepressants protected against the loss of hippocampal volume seen in patients with unipolar depression. These data suggest that appropriate prophylactic medications (antidepressants for unipolar depression, and lithium and the mood-stabilizing anticonvulsants and some atypicals for bipolar episodes) may help protect the brain in several different ways. First, to the extent that they are effective in preventing the occurrence of affective episodes, they would eliminate episode-related decreases in BDNF and increases in oxidative stress. Second, to the extent that they also directly increase BDNF and other neuroprotective factors and increase neurogenesis, they may help repair some of the deficits that occur in the illness (Fig. 1), or at least prevent them from progressing. Consistent with this viewpoint are the data that repeated lithium use reduces the risk of dementia in late life (Kessing et al. 2008).
Implications for Clinical Therapeutics
Not only is there a 10–15% lifetime incidence of suicide with the recurrent unipolar and bipolar affective disorders, but early mortality from a variety of medical causes is nothing short of stunning (Newcomer and Hennekens 2007). People with major mental illnesses (including unipolar depression and bipolar disorder) have life expectancies much lower than those of the general population. This difference has been estimated to be lowest but still 13 years in the state of Virginia and as large as 25–30 years in many of the western states of the United States (Colton and Manderscheid 2006). The vast majority of this shortened life expectancy is due to cardiovascular disease, and it is likely that many of the same mechanisms (Table 3) that contribute to abnormalities in the CNS also lead to deterioration of peripheral somatic systems as well (Kapczinski et al. 2008b).
Given these clinical and neurobiological observations, the fundamental view of mental illnesses should change. While they are now thought of as “mental” (i.e., too often connoting “abstract, imaginary, all in the mind, ephemeral”), they should instead be seen as recurrent, potentially progressive, and lethal medical illnesses of brain and body that require lifelong treatment and monitoring.
If we only used the currently available therapeutic tools more expeditiously, with earlier intervention and more sustained prophylaxis at stages (4) syndrome and (5) recurrence rather than the later stages 6 and 7 (Fig. 4), it is likely that we would make a dramatic impact on the course of the recurrent affective disorders. We know that early-onset recurrent unipolar depression and bipolar disorder are both poor prognosis factors for a negative outcome in adulthood (Perlis et al. 2004; Leverich et al. 2007). At the same time, we have observed that the duration of the delay to first treatment, which is inversely related to the age of onset of illness, is an independent risk factor for a poor outcome in adults with bipolar disorder (Post et al. 2010).
Thus, treating youngsters and adults with affective disorders earlier and more effectively may do much to convert their illnesses to more benign ones (Franchini et al. 1999; Post and Kowatch 2006), with the hope that this would ultimately change their adverse trajectory altogether (Fig. 4). While this remains to be directly demonstrated, from a clinical perspective, the adequate treatment of affective episodes themselves would have enormous personal, social, and economic benefits even if it did not alter the illness course in any fundamental way.
Epigenetic Mechanisms of Illness Progression
In the latter part of the past century, we have been able to trace both illness mechanisms and the actions of therapeutic agents to increasingly deeper levels of inter- and intra-cellular communication. Observations have moved from alterations in synaptic concentrations of neurotransmitters to second messenger systems, kinases, and a variety of transcription factors that affect DNA dynamics, and ultimately RNA and protein synthesis. The pathways involving changes in the nucleus have allowed us to conceptualize how acute stressors and short-term events in the environment may be translated into longer-term changes in neuronal and glial modulation in the weeks or months that an affective episode might last (Post 1992).
We are also aware of environmentally induced changes that can leave permanent residues, not only in the memories of events associated with posttraumatic stress disorder, but also in the set points for neurotransmitter biosynthesis, neurotrophic levels, and even neurogenesis (Coe et al. 2003; Roceri et al. 2004; Roth et al. 2009). Some of these longer-term to life-long neurobiological changes relevant to progressive behavioral alterations have recently been found to occur at the level of modifications in DNA and histone structure—what are called epigenetic changes (Tsankova et al. 2007). If DNA is wound tightly around histones, it is less available for transcription. Similarly, if DNA is methylated, the genes in that area of methylation are typically less likely to be transcribed. Stressors of sufficient quality and severity early enough in life are known to change DNA and histone methylation and histone acetylation, such that there is relative inhibition/repression or activation/induction of the genes near these structural modifications.
While these epigenetic changes have implied long-lasting and often negative outcomes, new data also reveal that this is not necessarily the case, and changes in DNA and histone methylation and acetylation are, themselves, potentially plastic and amenable to therapeutic alterations (Weaver et al. 2006; Tsankova et al. 2007; Roth et al. 2009). For example, the mood-stabilizing anticonvulsant valproic acid, in addition to its many other actions, is a histone deacetaylase inhibitor (Bredy et al. 2007). In this role, it prevents the removal of acetyl groups, which tend to yield looser winding of DNA around histones, thus increasing DNA transcription. Methyltransferases and methylation inhibitors likewise can be involved in direct alterations of methyl groups on the DNA itself (Roth et al. 2009).
Since most of our therapeutic agents are thought to work at the more transient levels of actions on neurotransmitters to transcription factors which modulate DNA promoter regions directly, they are effective in suppressing illness manifestations and affective episodes, but do not change the fundamental vulnerability to illness recurrence, and the need for life-long maintenance treatment (Post and Leverich 2008). The data that environmentally induced epigenetic modifications of DNA and histones once thought to be permanent or lifelong are reversible, raise the future possibility that initial and cumulative affective illness vulnerability could be altered by epigenetic manipulations in some fundamental way such that life-long prophylactic medicine in some cases would no longer be required.
Until we are able to modify or prevent some of the persistent environmentally induced epigenetic changes, it is likely that we will be dealing with an illness that retains many different potentialities for progression. Yet stressors, recurrent episodes, and substances of abuse each of which can propel illness progression are themselves subject to amelioration and prevention, and thus are current and urgent targets for more aggressive efforts at clinical therapeutics.
In the future, it is hoped that even these environmentally induced potentially accumulative vulnerabilities, to the extent that they are mediated by epigenetic alterations, may themselves be modifiable at this more fundamental level of gene regulation. For example, we previously noted that defeat stress results in downregulation of BDNF in the hippocampus. This occurs on the basis of dimethylation of histone K3H27 (Tsankova et al. 2007). When animals are treated with antidepressants, they are able to overcome this BDNF suppression by several different molecular and epigenetic mechanisms and thus prevent the defeat stress depressive-like behavior. However, the initial defeat stress-induced dimethylation of K3H27 is not altered by antidepressant treatment, and this leaves behind a residual memory mechanism or molecular scarring that could underlie the persisting increased vulnerability to subsequent adverse environmental events and abnormal behavioral outcomes.
This persisting dimethylation of H3K27 may prove to be an example of the type of epigenetic mechanism that so far we have not been able to directly reverse (Tsankova et al. 2007), but can only ameliorate indirectly with our current disease-suppressing therapeutic maneuvers that increase BDNF in the hippocampus by other pathways. We look forward to a future round of new therapeutic interventions that might directly modify some of these persistent and cumulative mechanisms of illness progression at the level of epigenetic modifications that could reverse some aspects of illness vulnerability.
The environmentally mediated changes in neurobiology provide a panoply of mechanisms (Table 3) for illness stage progression (Fig. 4), and even now many are potential targets for early intervention and secondary and tertiary preventions. As we move into the later part of this century, it is also conceivable that some of the genetic aspects of illness vulnerability coded in DNA sequences will also be amenable to alteration and revision and yield the possibility of primary prevention in those at highest risk.
- Antelman S (1988) Stressor-induced sensitization to subsequent stress: implications for the development and treatment of clinical disorders. In: Kalivas P, Barnes C (eds) Sensitization in the nervous system. Telford Press, Caldwell, NJ, pp 227–254Google Scholar
- Bearden CE, Thompson PM, Dalwani M, Hayashi KM, Lee AD, Nicoletti M, Trakhtenbroit M, Glahn DC, Brambilla P, Sassi RB, Mallinger AG, Frank E, Kupfer DJ, Soares JC (2007) Greater cortical gray matter density in lithium-treated patients with bipolar disorder. Biol Psychiatry 62:7–16CrossRefPubMedGoogle Scholar
- Benson B (2009) Interregional cerebral metabolic associativity in unipolar and bipolar disorder. Part II. Differential alterations in bipolar and unipolar disorders. Psychiatry Res 164:30–47Google Scholar
- Kraepelin E (1921) Manic-depressive insanity and paranoia. E.S. Livingston, EdinburghGoogle Scholar
- Krishnan V, Graham A, Mazei-Robison MS, Lagace DC, Kim KS, Birnbaum S, Eisch AJ, Han PL, Storm DR, Zachariou V, Nestler EJ (2008) Calcium-sensitive adenylyl cyclases in depression and anxiety: behavioral and biochemical consequences of isoform targeting. Biol Psychiatry 64:336–343CrossRefPubMedGoogle Scholar
- McElroy SL, Altshuler LL, Suppes T, Keck PE Jr, Frye MA, Denicoff KD, Nolen WA, Kupka RW, Leverich GS, Rochussen JR, Rush AJ, Post RM (2001) Axis I psychiatric comorbidity and its relationship to historical illness variables in 288 patients with bipolar disorder. Am J Psychiatry 158:420–426CrossRefPubMedGoogle Scholar
- Perlis RH, Miyahara S, Marangell LB, Wisniewski SR, Ostacher M, DelBello MP, Bowden CL, Sachs GS, Nierenberg AA (2004) Long-term implications of early onset in bipolar disorder: data from the first 1000 participants in the systematic treatment enhancement program for bipolar disorder (STEP-BD). Biol Psychiatry 55:875–881CrossRefPubMedGoogle Scholar
- Post RM (2008) Animal models of mood disorders: kindling as model of affective illness progression. In: Schachter S, Holmes G, Trenite D (eds) Behavioral aspects of epilepsy: principles, practice. Demos, New York, pp 19–27Google Scholar
- Post RM, Leverich GS, Kupka R, Keck R, McElroy S, Altshuler L, Frye M, Luckebaugh DA, Rowe M, Grunze H, Suppes T, Nolen W (2010) Early onset bipolar disorder and treatment delay are risk factors for poor outcome in adulthood. J Clin Psychiatry, in pressGoogle Scholar
- Post RM, Kauer-Sant’Anna M (2010) An introduction to the neurobiology of bipolar illness onset, recurrence, and progression. In: Yatham LN, Maj M (eds) Bipolar disorder: clinical and neurological foundations. John Wiley & Sons, LtdGoogle Scholar
- Post RM, Leverich GS (2008) Treatment of bipolar illness: a case book for clinicians and patients: WW Norton, Inc, New YorkGoogle Scholar
- Post RM, Miklowitz D (2010) The role of stress in the onset, course, and progression of bipolar illness and its comorbidites: implications for therapeutics. In: Miklowitz D, Cicchetti D (eds) Bipolar disorder: a developmental psychopathology approach. Guilford, New YorkGoogle Scholar
- Post RM, Post SLW (2004) Molecular and cellular developmental vulnerabilities to the onset of affective disorders in children and adolescents: some implications for therapeutics. In: Steiner H (ed) Handbook of mental health interventions in children and adolescents. Jossey-Bass, San FranciscoGoogle Scholar
- Post RM, Weiss S, Pert A, Uhde T (1987) Chronic cocaine administration: sensitization and kindling effects. In: Raskin A, Fisher S (eds) Cocaine: clinical, biobehavioral aspects. Oxford University Press, New York, pp 109–173Google Scholar
- Roybal K, Theobold D, Graham A, DiNieri JA, Russo SJ, Krishnan V, Chakravarty S, Peevey J, Oehrlein N, Birnbaum S, Vitaterna MH, Orsulak P, Takahashi JS, Nestler EJ, Carlezon WA Jr, McClung CA (2007) Mania-like behavior induced by disruption of CLOCK. Proc Natl Acad Sci USA 104:6406–6411CrossRefPubMedGoogle Scholar
- Torres IJ, Boudreau VG, Yatham LN (2007) Neuropsychological functioning in euthymic bipolar disorder: a meta-analysis. Acta Psychiatr Scand Suppl 434:17–26Google Scholar