Neurobiology of Suicide and Aggression

J. John Mann, M.D.

This chapter will address the question of violence and aggression by a consideration of physical acts that are directed by an individual toward other persons with the goal of causing them physical harm, or towards the self with the goal of suicide. Externally directed aggressive acts will be confined to those occurring in the context of an affect such as anger or fear, excluding acts towards others that are planned in the context of deliberate criminal behavior. The emphasis will be on a review of neurobiological mechanisms involved in such self-directed or outwardly-directed aggression and the implications of these mechanisms for possible pharmacological treatment approaches. The reason for this emphasis is to be consistent with the overall theme of the volume, that is neuropsychopharmacology. Social, cognitive, behavioral and other psychological mechanisms are also of fundamental importance in understanding violence and aggression (48), but careful treatment of these important areas is beyond the scope of this chapter.

With regard to externally directed aggression, there is a considerable animal literature that will be reviewed and integrated with clinical findings. With regard to suicidal behavior, there is no significant animal literature, and the review will therefore concentrate on human studies. There is, however, a relationship between internal and externally directed aggression, based partly on the observation of a statistically significant association of both types of aggression within the same individual. Therefore, externally directed violence will not be discussed independently of internally directed violence or suicidal behavior, and common and different features will be noted where relevant.

Suicide is the 8th leading cause of death and homicide is the 10th highest cause of death, based on data for 1991 reported by the National Center for Health Statistics. According to data from the Federal Bureau of Investigation, there were 23,760 homicides in the United States in 1992. Poverty and drug abuse are associated with a higher rate of violence and may contribute to the elevated homicide rate in minorities such as young African-American males. Over 90% of murderers are males. Violent crime by females appears to be increasing (19)

In contrast, suicides occur most frequently among young white males. The Centers for Disease Control report that in 1990 there were 30,906 suicides in the United States, of which 24,724 occurred in males and 22,448 in white males. The reason for these racial and sex differences in the proportions of suicide and homicide is unclear but may provide clues as to how to favorably alter the rates of both problems. For example, males make up a majority of both murderers and suicide victims. The Epidemiological Catchment Area (ECA) Study found that violent behavior in the year prior to interview was twice as prevalent in males compared to females (71). Studies in noncriminal populations find that males are more aggressive than females. Understanding these behavioral differences in males and females may also help to identify ways of reducing serious aggression.

In many cases the violence appears to be clearly directed inwardly or outwardly, however a closer examination of the data indicates that this separation is not a strong as it may first appear. First, there are cases involving homicide followed by suicide (16). In these cases, there are a disproportionate number of multiple homicides before the perpetrators kill themselves. At a less extreme level, examination of murderers shows a significant rate of previous suicide attempts or at least self-destructive acts that have threatened the life of the individual and may be interpreted as suicidal acts. About 30% of violent individuals have a history of self destructive acts. Conversely, 10–20% of suicidal individuals have a past history of violent behavior towards others. As a group, individuals presenting with an episode of major depression or a psychotic illness and having a history of a past suicide attempt are distinguished from individuals with the same psychiatric illness and no lifetime history of a suicide attempt by having a higher level of lifetime externally directed aggression and impulsivity. Thus, suicide is a less common form of aggression that is self-directed and, like externally directed aggression, it is a consequence of a lower threshold for acting on strong emotions. This lower threshold can also be termed as the propensity for impulsivity.

The problem of violence extends beyond the extremes of homicide. Whereas in 1984 the rate of murder was 7.9 per 100,000 people per year, the rate of aggravated assault was estimated at 290 per 100,000 per year, and the rate of forcible rape was 35.7 per 100,000 per year. Thus, the extent of the problem of aggression and violence is enormous and has led to considerable controversy in terms of seeking methods for controlling violence, including legislating gun control, increasing prison capacity, mandatory sentences for violent offenses, attempts to control drug-related sales and purchases, lobbying to reduce violence on television and films, and public education. There is no evidence that these measures are being pursued either to the full extent that is possible, or that the attempts to pursue these avenues have been particularly successful. The type of externally directed violence addressed in this chapter may well account for only a small fraction of the type of violence that has placed the nation in crisis. Psychiatric illness is associated with a minority of violent acts (48), whereas psychiatric illness is associated with suicide in over 90% of cases. Thus, this chapter addresses mechanisms that are involved in a minority of cases of violence against other persons but are applicable to the vast majority of cases of suicide.

This chapter is organized into two major parts. The first part addresses externally directed aggression and begins with animal studies of aggression, followed by clinical studies of serenics or antiaggression pharmacotherapy, and finally the neurobiology of human aggression. The second part covers suicidal behavior, beginning with the neurobiology of completed suicide and then deals with attempted suicide. The reader is referred to the chapter by B. Eichelman (22) in the 1987 ACNP Volume "Psychopharmacology: A Third Generation of Progress." Because of the reference limit in this volume, all relevant aggression references in that earlier chapter are referenced via the chapter in the text.

Animal models of aggression are generated in two major ways. The first involves development of models of aggression through manipulation of the brain pharmacologically or by lesioning in order to induce aggressive states. A second approach utilizes behavioral manipulation to produce a model of aggressive behavior. These animal models of aggression have been used to test a variety of pharmacological agents that may have potential in the treatment of human aggression (for reviews see refs. 22, 32, 52).

There are two major types of animal models of aggression: offensive and defensive aggressive behavior. A third domain of behavior that has been associated with aggression or aggressivity is dominance. It is assumed that more dominant animals are more aggressive. This is an oversimplification, since dominant animals may intimidate other animals in their group without actual physical contact. Models of offensive aggression include isolation-induced aggression, the resident-intruder paradigm (or territorial aggression) and maternal aggression. In isolation-induced aggression, male mice are isolated for two to six weeks and then placed together for fighting. The latency, duration and intensity of the fighting are measured and then effects of pharmacological intervention can be assessed. In territorial aggression models, a male rat or mouse is housed with a female. When another male is introduced into the territory, heavy fighting may be triggered. Lesions of the hypothalamus of male or female rats can produce similar behavior. Another model involving maternal aggression is created when a mouse is killed by a rat, referred to as muricide. A related model involves philicidal behavior, which involves pup killing. In most of these models, including shock-induced aggression, decreasing serotonergic activity results in an increase in aggression and increasing serotonergic activity results in a decrease in the aggressive behavior. These data have been reviewed extensively by other investigators (14, 18, 22, 23, 32, 69, 76).

Defensive aggression models include the characteristic of the lack of an active approach and usually no wounds are made on the attacker (22, 52). Examples of this model are foot shock or pain induced aggression, and the rat resident-intruder paradigm or maternal aggression in response to an intruder. Therefore, this kind of defensive behavior model reflects a different form of aggression than the offensive aggression models. However, this behavior may also be manipulated by serotonergic activity. Defensive aggression models have not been as extensively used in pharmacological drug development, perhaps because they are viewed as less "pathological."

In recent years, two developments have reawakened interest in the study of aggression in animal models. The first was a series of studies carried out in nonhuman primates; the second involved the cloning of multiple serotonin receptor subtypes whose specific role in mediating aggression can be studied.

With regards to the serotonin receptor subtypes of interest, it has been known for a number of years that 5-HT2 receptor antagonists, which include most antipsychotic drugs, may have antiaggressive properties. Of considerable interest is the recent observation by Hen and colleagues (28) that deleting the 5-HT1B gene in mice produced a decrease in latency to attack and an increase in attack behavior using the intruder-resident mouse model. This raises the possibility that the 5-HT1B receptor mediates the antiaggressive effects of the transmitter serotonin.

The nonhuman primate studies have demonstrated a relationship between low levels of serotonin and aggressive behavior and potentially either a state- or trait-dependent relationship with dominance hierarchy. CSF 5-HIAA was lower in dominant monkeys, associated with aggression and impulsivity and platelet serotonin content was higher in dominant monkeys in a state-dependent fashion (60). A high cholesterol diet has been shown to raise CSF 5-HIAA levels and increase the prolactin response to serotonin in monkeys. This enhancement of serotonergic function is accompanied by decreases in levels of overt aggression and more socialization. Thus, cholesterol appears to be one mechanism whereby the serotonergic system can be manipulated independently of the noradrenergic and dopaminergic systems, and these manipulations appear to have behavioral consequences which involve overt aggression and degree of social contact with peers. The serotonergic system has been found to be under significant genetic control both in monkeys (59) and in humans (47, 54), and may represent a biochemical trait with behavioral trait consequences. Peer-raised monkeys appear to have reset their serotonergic activity at a lower level than maternally raised monkeys, indicating that maternal deprivation can have enduring deleterious effects on the serotonergic system and, thereby, behavior.

Assessment of the efficacy of pharmacological agents for the treatment of aggressive and violent behavior requires a distinction to be made between specific antiaggressive effects and nonspecific effects such as the induction of sedation, hypotension, extrapyramidal symptoms such as hypokinesia and other nonspecific effects. Drugs that have been widely used as serenics include benzodiazepines, anticonvulsants, antipsychotics and b-adrenergic receptor (BAR) antagonists, and all have such nonspecific effects. Several published reviews discuss the efficacy of drugs in animal models of aggression (22, 52). This chapter will focus on human studies.

Antipsychotic medications have been used for many years as antiaggressive drugs (22). However, studies in animals suggest that their action may be more nonspecific than commonly appreciated. While we and others have found a relationship between dopaminergic activity and dominance, it is more difficult to find direct relationships of dopaminergic activity with aggressivity. Many antipsychotic drugs are not only dopamine receptor antagonists, but they also have sedating effects, antiserotonin effects and can reduce blood pressure through a1-adrenergic blockade. Newer antipsychotic drugs that have more selective effects for dopamine receptor subtypes such as the action of clozapine on D2 and D4 receptors, and the action of clozapine and other antipsychotics on 5-HT2 receptors, raise the possibility of more specific antiaggression agents. Controlled clinical trials of these agents are warranted. Lithium carbonate is useful in the treatment of certain types of aggressive disorders (22), an effect attributed to its serotonin enhancing action. Anticonvulsants have been used to treat aggressive behavior (22), initially based on the observation that certain types of seizure disorders were associated with aggressive behavior. This raised the possibility that subictal discharges could be associated with behavioral discontrol or aggressivity. However, the data supporting the usefulness of anticonvulsants are relatively weak outside the context of a clearcut seizure disorder. Benzodiazepines have also been used for treating aggressive behavior (22) and have both sedative and anticonvulsant properties via enhancement of GABAergic function. Beta-adrenergic receptor antagonists have been used for treating acute aggressive outbursts. Anti-androgen drugs also have been used to treat aggression (22). Most of the studies have been carried out in Europe, and double-blind, placebo-controlled studies to determine whether these approaches are efficacious are lacking.

Conversely, naturalistic and some controlled studies have demonstrated that alcohol, cocaine, and amphetamines can potentially increase aggressive behavior (22, 25, 58). Drugs such as amphetamines release norepinephrine, dopamine and serotonin. Drugs such as cocaine are monoamine reuptake inhibitors, particularly of dopamine and norepinephrine. Enhancement of transmission in dopaminergic and noradrenergic systems may increase aggressivity. Paradoxically, acute alcohol administration also causes release of serotonin. All of these agents share the common feature of a depletion state following an initial excess of transmitter. This depletion state appears to be associated with depression. One formulation of these results is that the pronounced release of dopamine and perhaps norepinephrine leads to the increase in aggression. That release overwhelms any restraining or modulating effect of serotonin release. Another, related formulation is that significant aggression results from the use of these agents in individuals who are already vulnerable due to brain damage or other neurobiological predispositions. In such individuals, drugs are only one of an array of potential triggers of aggression (25).

This review will focus on the potential role of the noradrenergic, serotonergic, and dopaminergic systems in aggression. In addition, the role of brain injury and genetics in aggression is discussed.

Human studies of the noradrenergic system have involved two major methodologies. The first is the neuroendocrine strategy, such as the use of growth hormone response to the a2-adrenergic agonist clonidine. Siever and Trestman (68) showed that clonidine-stimulated growth hormone responses were increased in patients with personality disorders, and that this increase correlated with sensation-seeking and risk-taking behaviors. They did not find a correlation with overt aggression or impulsivity. This finding of an enhanced growth hormone response to clonidine contrasts with reports by the same group and others of a blunted growth hormone response in depressed patients which persists into the well state. Thus, opposite findings were made in depression and aggression. A second strategy has involved the measurement of cerebrospinal fluid levels of 3-hydroxy, 4-methoxyphenylglycol (MHPG). Brown and colleagues (11) reported a positive correlation between a history of aggressive behavior in military personnel, personality disorder and levels of CSF MHPG. Consistent with these findings are reports that the b-adrenergic antagonist propranolol may be effective in the treatment of episodic aggressive behavior. This kind of behavior is frequently associated with head injuries.

As indicated previously, some evidence from animal studies indicates that the noradrenergic system is involved in aggression. Stimulation of the amygdala produces sham rage (61), and this behavior was associated with a fall in brainstem and brain levels of norepinephrine. Presumably, the drop in levels of norepinephrine is a reflection of norepinephrine release. Other animal models support a role for increased norepinephrine release in aggression. What is paradoxical is that certain animal models of aggression, such as shock-induced fighting, are induced more readily with lowered brain norepinephrine levels.

The most consistent data address the role of the serotonergic system in human aggression. Brown and colleagues (11),in a study of subjects with personality disorders, found that CSF 5-HIAA was inversely correlated with clinician or self-reports of lifetime aggression. Other investigators reported negative correlations of CSF 5-HIAA with irritability (14), hostility (14), impulsive homicide (38), or fire setting (75), maternal aggression, and with self-rated behavioral difficulties during childhood. Recidivism of murderers has also been showed to be correlated with low levels of CSF 5-HIAA (75). Neuroendocrine studies using the agent fenfluramine have found an inverse correlation between the prolactin response to fenfluramine and irritable, impulsive aggression in patients with personality disorder (14), but the same investigators were unable to demonstrate such a relationship in depressed patients. We recently observed this relationship in a larger sample of depressed patients, as well as in a nonpsychiatric group, but mostly in males and postmenopausal females. There is also evidence for a similar negative correlation between the prolactin response to the direct 5-HT2A/5-HT1C receptor agonist m-chlorophenylpiperazine and impulsive aggression. There is also a significant positive correlation between the prolactin response to mCPP and to fenfluramine, suggesting that a significant part of the variance in the prolactin response to fenfluramine is accounted for by 5-HT2A/5-HT1C receptor responsivity (15). More recently, a correlation has been found between the genotype for a polymorphism of an intron (non-coding section) in the gene for tryptophan hydroxylase and levels of CSF 5-HIAA in impulsive aggressive individuals (49) This suggests a link between a gene involved in the serotonergic system and impulsive aggression. Since the serotonergic system is under significant genetic control (47, 54), and there is a genetic contribution to aggressivity (46, 66), part of the basis for a genetic vulnerability to aggression may be explained by low CSF 5-HIAA as an inherited trait.

Head injuries involving the prefrontal cortex are associated with aggressive behavior. Seizure disorders of various types, particularly those involving the temporal lobe, are also associated with episodic and impulsive aggression. These studies suggest that organic injuries to the brain, particularly prefrontal cortex and perhaps certain limbic areas, somehow interfere with mechanisms involving inhibition of aggressive behavior. The failure of such inhibitory mechanisms results in an increased likelihood of aggression and violence.

There is a genetic component involved in aggression (46, 66). The mechanism of this genetic component remains uncertain. We have already suggested that a reduced level of serotonin function may be one way in which genetic effects could predispose an individual to aggressive behavior. Since there is evidence that the dopaminergic system is also under genetic control (59), and greater dopaminergic activity has been associated with dominance and aggressivity, it is conceivable that genetic factors could also result in increased dopaminergic activity, leading to greater aggressivity. Few direct studies of the dopaminergic system in aggressive individuals have been done. Pharmacological data concerning the effects of amphetamine (22) and the antiaggressive effects of dopamine antagonists (22) support the hypothesis that increased dopaminergic activity may underlie some forms of aggression. Two other observations are relevant. Individuals with additional Y chromosomes are more common in violent, criminal populations. As males are more aggressive than females, there may be a role for the Y chromosome in aggression. Testosterone may play a role in mediating aggression, as well. A recent study (12) reported that a point mutation in the eighth exon of the gene on the X chromosome for monoamine oxidase (MAO) A causes a failure in transcription and results in an absence of MAO A in affected males. In this single family, the absence of MAO A was associated with lower intelligence and impulsive, aggressive behavior in males only. The lack of MAO A activity would result in reduced levels of 5-HIAA, HVA and vanillyl-mandelic acid (VMA) and increased levels of NE, DA and serotonin (12).

It is highly probable that multiple transmitters are involved in modulation of aggressive behavior. The serotonergic system does have some inhibitory effects on the dopaminergic system, and reduced serotonergic function may thus result in increased dopaminergic function. Such a relationship may explain the potential co-existence of these two transmitter system abnormalities.

There are approximately 31,000 suicides per year in the United States. The majority of these suicides occur in the context of a major depressive episode. Schizophrenia, personality disorders, alcoholism, and substance abuse are other psychiatric disorders frequently associated with suicide. Over 90% of suicides suffer from a psychiatric disorder, and it is uncommon for suicide to occur in a individual who is not psychiatrically ill. Conversely, many patients with psychiatric disorders may experience suicidal ideation but do not commit suicide. This raises the following important question: what distinguishes psychiatrically ill individuals who commit suicide from those who do not? A second related question is whether there is a common factor or set of factors across psychiatric diagnoses that predispose individuals to commit suicide, and whether there are other factors related to suicide risk in psychiatric disorders that are specific for each disorder. We will discuss in some detail the result of studies in postmortem brain tissue from suicide victims and data from studies comparing suicide attempters with psychiatric controls.

There are no models of suicidal behavior in animals. One approximation is self-injurious behavior in monkeys. This model has been reviewed in detail by Kraemer and Clarke (34). Essentially, isolation has been found to increase the probability of self-injurious behavior during adolescence and adulthood. There is evidence of abnormalities in both the serotonergic and noradrenergic systems in self-injuring monkeys. This behavior tends to get worse as the monkey matures and is principally modified by pharmacological agents acting on the serotonergic system. Kraemer and Clarke (34) report that this response to improvement with serotonergic agents seems to be somewhat individualized. The relationship of self-injurious behavior to the human condition as a model of suicide attempts is questionable, and it may be a better model of Lesch-Nyhan Disorder (51) and other such unusual biochemical abnormalities in man. Self-injurious behavior may also be displayed by certain patients with personality disorders or cognitive deficits. Such behavior is distinguished from suicidal acts by and by the lack of interest in suicide. Its repetitive, ritualistic pattern and compulsive quality.

In the ever-increasing literature describing postmortem studies of suicide victims, the majority of the work has concentrated on the serotonergic system and to a lesser extent on the noradrenergic system. Other neurotransmitter systems, including the GABAergic system, cholinergic system, dopaminergic system and peptide modulators and transmitters, have been studied to a far lesser degree, thus preventing definite conclusions. This review will be confined to the serotonergic and the noradrenergic systems, where there is the greatest amount of information available. Transmitter turnover and receptor studies and their relationship to the type of suicidal behavior (violent vs. nonviolent) or the underlying psychiatric disorder will be themes that guide this review.

There have been approximately 14 studies of the concentration of serotonin (5-HT) and its major metabolite 5-hydroxyindoleacetic (5-HIAA) in brain tissue from suicide victims (see Mann et al. 1994 {43} for a review). Five of the seven studies of suicide victims found reductions in either 5-HT or 5-HIAA in the brainstem. In contrast, studies of the prefrontal cortex found a reduction in 5-HIAA levels in only three of eight studies, and no study found a reduction in serotonin. Examination of other brain regions found reductions in 5-HT or 5-HIAA in four out of seven studies. The degree of reduction in 5-HT or 5-HIAA was similar in depressed patients, alcoholics, schizophrenics, and patients with personality disorders (35). Similarly, use of a violent suicide method was not associated with a greater degree of decrease in serotonin or 5-HIAA, compared with nonviolent suicides (41). Therefore, the method of suicide appears to be unrelated to the biochemical findings.

In summary, there appears to be evidence for a modest reduction in levels of 5-HT and 5-HIAA in the brainstem of suicide victims, independent of psychiatric diagnosis. It is an open question as to whether 5-HT and 5-HIAA are altered in the prefrontal cortex or other brain regions. Postmortem studies of monoamines and metabolites suffer from the inability to distinguish the transmitter levels in the high-turnover, releasable pool from the lower turnover, storage pool. A second difficulty is the high rate of degradation postmortem. What lends these findings of reduced brainstem serotonin and 5-HIAA more credibility is that most studies of CSF 5-HIAA find lower levels in suicide attempters.

A more stable postmortem index of serotonin function is the receptor protein. The serotonin receptor that has received the greatest attention is the serotonin transporter. There have been at least 15 published studies of serotonin transporter binding or [3H]imipramine binding in suicide victims (see Mann et al. for a review; ref. 43). Four of these studies reported a decrease in imipramine binding. Fewer of the studies that used more specific ligands than [3H]imipramine, or used a more selective displacer than desipramine, found a reduction in binding, for example a study by Laruelle (36), who used [3H]paroxetine combined with clomipramine as a displacing agent. It is possible that the reductions reported in earlier studies may in fact have involved a binding site other than the physiologically relevant transporter site. The functional role of this other binding site is unclear. The nontransporter site is found in greater numbers than the transporter site itself, and further research is required to determine whether or not it has a physiological role. Another factor to consider is the brain region that is being studied. For example, an autoradiography study of suicide victims by Gross-Isseroff et al. (27) found regions of unchanged, increased and decreased 3H-imipramine binding. Our own studies using autoradiography show that the reduction in cyanoimipramine binding to the transporter site was confined to the ventral prefrontal cortex. This suggests that, in addition to ligand specificity, there may indeed be a reduction in transporter binding, but the reduction appears confined to ventral areas of prefrontal cortex and not dorsal prefrontal cortex that was commonly studied previously.

The 5-HT2A serotonin receptor is one of the earliest identified postsynaptic serotonin receptors in cortical tissue. In 1983, we first reported a 44% increase in [3H]spiroperidol binding to the 5-HT2A receptor, as defined by mianserine in prefrontal cortex (43). We replicated the same result in two other series of brains in 1986 and 1990 (1, 42). The latter study by Arango et al. (1) was carried out using both membrane binding with [¹²⁵I]LSD and autoradiography. That study not only confirmed an increase in binding in the prefrontal cortex but also found that it involved a greater number of binding sites with no difference in the K_D. Moreover, the degree of difference between the suicides and the controls appears to be greater in prefrontal cortex than in temporal cortex, suggesting regional specificity of the suicide effect. Of the five studies that did not replicate the result (13, 27, 53), four came from a single research group and one was an independent autoradiography study by Gross-Isseroff et al. 1990 (27). However, other investigators have replicated this finding, including Arora and Meltzer 1989 (66), Laruelle 1983 (36), and Yates 1990 (78). The latter studies were conducted with a series of depressed patients who were dying of natural causes. Thus, there is evidence both for and against this finding. Further work is needed to map the distribution of change in 5-HT₂ receptors in suicide victims throughout the prefrontal cortex, as well as in other cortical brain regions.

At least five studies have been published of 5-HT1A receptors in suicide victims (see Mann et al. {43} for a review). In prefrontal cortex, the 5-HT1A receptor is predominantly postsynaptic. Of the five published studies, two reported an increase in this receptor in suicide victims and three did not. One factor that may be relevant for explaining these discrepant results is that the two studies that found an increase in 5-HT1A receptor binding (Arango et al. 1991 {3} and Joyce et al. 1993 {31}, both found this increase to be confined to very discrete brain regions. Thus, techniques such as autoradiography, that can map the regions likely to show receptor binding changes, are essential for detecting highly localized alterations.

Other serotonin receptor subtypes have barely begun to be investigated, and studies are ongoing of 5-HT1C, 5-HT1B and 5-HT1D receptors in suicide victims. Overall, the preponderance of data suggests that there are alterations in the serotonin system in suicide. The use of techniques such as autoradiography, coupled with gene expression studies, will help to clarify the range and extent of the receptor and transmitter changes, as well as identifying where in the brain these changes are most pronounced.

Too few postmortem studies of cortical or brainstem norepinephrine concentrations have been carried out in suicide victims to draw firm conclusions. No change in NE concentrations was found in brainstem, and increased NE in the cerebral cortex provided little guidance as to NE levels available for release or what the level of noradrenergic activity might be. Other investigators examined the cerebrospinal fluid (CSF) of suicide attempters as a way of avoiding the post-mortem degradation of neurotransmitter and obtaining an index of neuronal activity. Except for a couple of studies (4, 5), the majority (11, 57, 64, 65, 67) did not find reduced concentrations of the principal NE metabolite 3-methoxy, 4-hydroxyphenylglycol (MHPG) in the CSF of suicide attempters. Excretion of MHPG may be reduced in the urine of suicide attempters (4, 5).

Increased binding to b-adrenergic receptors in the cerebral cortex of suicide victims, compared with controls, was reported by some investigators (1, 10, 42), but not by others (20, 39, 70). We now have evidence that there is decreased b1-adrenergic binding, which suggests that the increased binding reported with non-selective ligands may reflect increased b2-adrenergic binding. Although few studies have examined the a-adrenergic receptor subtypes, a1-adrenergic and/or a2-adrenergic receptor binding in cerebral cortex of suicide victims has been reported as increased (2, 45) and decreased (26). Additional studies are warranted, particularly since different ligands were used by different studies to define the a2-adrenergic receptor subtype.

Fewer locus coeruleus noradrenergic neurons, increased tyrosine hydroxylase and a2-adrenergic autoreceptor binding suggest that noradrenergic overactivity may have resulted in depletion. Perhaps the severe stress of the pre-suicide phase causes noradrenergic overactivity that, combined with fewer noradrenergic neurons, led to noradrenergic depletion.

These findings are a snapshot of brain function at the moment of suicide and therefore include the effects of genetics, development and early life processes, the associated psychiatric disease, as well as any treatment, environmental stresses and the artifacts of the postmortem delay. However, several other questions that are critical to the interpretation of our existing knowledge can be asked and should serve to direct future studies. Are the serotonergic changes genetic in origin and the noradrenergic changes state-dependent? Is there a highly specific and sensitive serotonergic trait marker which will identify the person at risk for suicide? Identification of the brain systems most affected or predictive of suicide may then provide the opportunity for effective pharmacological intervention. Much remains to be learned through further systematic postmortem neurochemical studies of tissue from individuals where clinical information is available.

Reduced levels of CSF 5-HIAA appear associated with higher rates of a history of planned, nonimpulsive suicide attempts, as well as higher rates of suicide attempts resulting in more medical damage. It was originally reported that CSF levels of 5-HIAA were distributed bimodally in depressed patients (6), and that the group with lower levels of CSF 5-HIAA was distinguished by a higher rate of serious suicide attempts and a higher rate of future suicide. However, suicide intent and medical damage were not quantified in that study. Since that report, many other studies have been published examining the relationship of CSF 5-HIAA to suicidal behavior in depressed and other psychiatric populations. Ten previous studies (4, 6, 21, 30, 40, 55, 73) found lower levels of CSF 5-HIAA in patients with a major depressive disorder who had also carried out a suicide attempt, compared with depressed patients who had not attempted suicide. Five studies did not find lower CSF 5-HIAA in suicide attempters (63, 67, 77). Four of seven studies found lower CSF 5-HIAA levels in schizophrenics who attempted suicide (17, 37, 50, 57, 64, 74, 77). However, all six studies (7, 11, 24, 72) of personality disorders and violent criminals (if the study of a mixed group of psychiatric patients by Banki & Arato {7} is included) found lower CSF 5-HIAA in suicide attempters. Three studies that did not find reduced CSF 5-HIAA levels in depressed patients who attempted suicide, had included a significant number with bipolar disorder. One study found lower CSF 5-HIAA levels in association with suicidal behavior in unipolar but not in bipolar depressed patients (4). However, a second study, where the depressed population comprised about 50% bipolar cases, did find reduced CSF 5-HIAA in the attempters (7). Thus, there is a proponderance of studie in the literature reporting that CSF 5-HIAA is lower in depressed or schizophrenic patients who have made a suicide attempt. This finding does not appear to be restricted to by diagnostic group, as it applies to both schizophrenia and depression. It now appears that suicidal behavior involving more planning and resulting in greater medical damage is significantly associated with lower CSF 5-HIAA levels, and therefore the type of suicidal behavior exhibited by the study population may potentially explain some of the differences in results in published studies.

With regard to suicidal behavior and CSF levels of HVA or MHPG, no consistent pattern has emerged in the literature. Some, but not all, studies report lower levels of CSF HVA in attempters (4, 63, 72). In addition, there is disagreement as to whether this relationship holds true only for major depressive disorder or whether it is also present in other diagnostic groups (63). Cerebrospinal fluid MHPG appears to have a unimodal distribution in affective disorders, with levels tending to be elevated relative to controls (33). One study found a complex relationship between CSF MHPG and suicidal behavior (4), but our study found no such correlation. We found that, unlike CSF 5-HIAA, CSF MHPG and HVA are not statistically significantly related to planning or medical damage, indicating a biochemical selectivity for this biobehavioral correlation.

It has also been hypothesized that the dimension of the suicidal behavior related to low CSF 5-HIAA is the degree of violence of the attempt, as opposed to some other aspect of suicidal behavior such as suicidal intent. Only three out of eight studies with available data reported that violent suicide attempts are associated with lower levels of CSF 5-HIAA. Therefore, the weight of the evidence is against this hypothesis. It is relevant to note that we have reported (44) that availability of method is a major factor determining choice of suicide method. Thus, selection of a violent method for suicide may not be biologically determined. If so, then another dimension of suicidal behavior may be biologically determined.

Although impulsive violence correlates with reduced serotonin activity (38, 75), previous studies did not directly examine the relationship between the degree of impulsivity involved in the suicidal act and CSF 5-HIAA. It has been reported (8) that Beck Suicide Intent Scale scores are higher than in patients who completed suicide, compared with patients who have attempted suicide, and this scale assigns higher scores to planning. Moreover, Suicide Intent Scale scores are higher in patients who reattempt suicide in the future (8) and in those who ultimately complete suicide (8). Degree of planning is associated more frequently with greater medical damage, and, from a behavioral standpoint, suicide attempters who plan their attempts resemble completed suicides.

With respect to the observation that greater planning is associated with more severe medical damage, failed suicides appear to be related to completed suicide, not only in terms of behavioral aspects but also in biological measures of serotonin function. Further evidence in favor of such a category of suicide attempters is provided by our findings that the degree of planning correlated with the degree of medical damage, such that more highly planned suicide attempts result in greater medical damage. Others have also reported that the degree of medical damage correlates with the score on the Suicide Intent Scale (56). It appears that planning for a suicide attempt is somewhat independent of impulsivity. Suicide attempts, whether planned or not, are associated with a history of greater lifetime impulsivity and externally directed aggression. Lifetime impulsivity correlates with the number of past suicide attempts in individuals with major depression and comorbid borderline personality disorder. Figure 1 illustrates the relationship of acute psychopathology, traits and suicidal or aggressive behavior.

Levels of CSF 5-HIAA do not correlate with the recency of the suicide attempt, but only with the characteristics of the suicide attempt. Data indicate that CSF 5-HIAA levels are stable in individuals tested on multiple occasions (29, 47) and can predict future suicide and attempts. This suggests that CSF 5-HIAA reflects a biochemical trait that may be determined genetically or as a result of developmental effects (54). Some studies indicated that suicide risk is at least partly genetically determined or familial (62). Thus, CSF 5-HIAA may be a biochemical trait related to the threshold for suicidal behavior and can be measured even if the suicide attempt was carried out some time earlier. Pharmacological elevation of serotonergic activity during periods of higher risk may protect patients from suicide or serious suicide attempts and perhaps mitigate the degree of medical damage in the event of a suicide attempt.

Aggression can be self-directed (suicidal acts) or externally directed. Aggression of both types tends to coexist in the same individual. The neurotransmitter correlates of aggression include decreased serotonergic activity and increases noradrenergic and dopaminergic activity. Reduced serotonergic activity correlates with planned and highly medically damaging suicidal behavior. Genetic factors and acquired brain injuries also contribute to violent behavior and suicide risk. Studies of the neurobiology of aggression must seek to distinguish state and trait factors and develop methods for detecting high risk patients who are candidates for pharmacotherapy. The neurobiological component that determines the predominant direction of aggression, namely inward (suicide) or outward, remains to be identified. Development of drug treatment for aggression or suicide risk requires a greater emphasis on controlled clinical trials in high risk populations. Identification of new serotonin and dopamine receptor subtypes offers the possibility of more specific serenic and antisuicidal medications.

This work was supported by the MHCRC for the Study of Suicidal Behavior MH46745.

published 2000